* master:
  go/zodb: Fix ΔTail.LastRevOf based on WCFS experience

- the only valid range for at is [tail, head]. Don't try to return anything
  meaningful for queries outside of this range and just panic instead. This is
  consistent with SliceByRev, which also panics on invalid query, and it is also
  consistent with semantic model that ΔTail is a vector with data keyed by tid in
  range (tail, head]: if key is out of vector range, access to the vector should
  panic, isn't it?

- instead of returning revision of minimum entry on exact=n, always
  return (tail, exact=n) in that case.

The change in behaviour is consistent with ΔFtail and ΔBtail from WCFS
and is needed for ΔFtail to function correctly:

https://lab.nexedi.com/kirr/wendelin.core/blob/22f5f096/wcfs/internal/xbtree/δbtail.go
https://lab.nexedi.com/kirr/wendelin.core/blob/22f5f096/wcfs/internal/zdata/δftail.go

* master:
  go/zodb: Prevent newly-created not-yet committed objects to go to ghost on deactivate
  go/zodb: Fix PActivate not to panic after an error
  go/zodb/fs1tools: Verify
  go/zodb/fs1tools: Dumper += DumpEndOK
  go/zodb: tests: Add a couple of missed t.Helper()
  go/zodb: tests: Teach tDB.Commit to return committed TID

When object is just created, it is not yet assigned an OID, but can be
reachable from other objects. The code that processes transaction can
reach to that new object and try to PActivate/PDeactivate it. And
currently PDeactivate will drop the object state completely.

Another example of object without an OID is Bucket embedded into a Tree
object. There, the code that scans the tree can reach to that bucket and
try to activate/deactivate it, leading, again, to dropping state of that bucket.

-> Fix it.

Persistent.PActivate used to panic when called the second time, if the
first time it hit an error. WCFS hit this in practice via object, that was
previously accessed and pinned into the cache, but later deleted in the
storage.

-> Fix PActivate to reset .loading on an error, so that next time PActivate is
called, it tries to trigger load again instead of panicking. Change doload
criteria from

	state==GHOST && refcnt==1

to

	state==GHOST && loading==nil

because now, after failed PActivate, refcnt can be != 0, if there are several
other PActivate calls that were waiting for the failed PActivate but did not
yet woke up.

Here is how added test fails without the fix:

    --- FAIL: TestActivateAfterDelete (1.65s)
    panic: t.zodb.MyObject(0000000000000065): activate: need to load, but .loading != nil [recovered]
            panic: t.zodb.MyObject(0000000000000065): activate: need to load, but .loading != nil

    goroutine 10085 [running]:
    testing.tRunner.func1.2(0x649020, 0xc000520660)
            /home/kirr/src/tools/go/go/src/testing/testing.go:1143 +0x332
    testing.tRunner.func1(0xc0001cb080)
            /home/kirr/src/tools/go/go/src/testing/testing.go:1146 +0x4b6
    panic(0x649020, 0xc000520660)
            /home/kirr/src/tools/go/go/src/runtime/panic.go:965 +0x1b9
    lab.nexedi.com/kirr/neo/go/zodb.(*Persistent).PActivate(0xc0001184d0, 0x6e8360, 0xc00019ac90, 0x0, 0x0)
            /home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/go/zodb/persistent.go:191 +0xce5
    lab.nexedi.com/kirr/neo/go/zodb.TestActivateAfterDelete(0xc0001cb080)
            /home/kirr/src/neo/src/lab.nexedi.com/kirr/neo/go/zodb/persistent_test.go:786 +0x72c

Add utility to verify FileStorage data for consistency.
To verify we just need to iterate through all records, because
FileStorage driver performs all consistency checks by itself.

Mimic normal output to be the same as in fstest from ZODB/py.
Example runs of fstest.py and `fs1 verify` on a broken file:

    $ python ~/src/wendelin/z/ZODB/src/ZODB/scripts/fstest.py -v 1.fs
             4: transaction tid 0x03e044f6448c8022 #0
           213: transaction tid 0x03e044f646e044bb #1
    1.fs has data records that extend beyond the transaction record; end at 466

    $ fs1 verify -v 1.fs
             4: transaction tid 0x03e044f6448c8022 #0
           213: transaction tid 0x03e044f646e044bb #1
    2021/05/24 12:43:37 fsverify: 1.fs: 1.fs: transaction record @355: -> (iter data): 1.fs: data record @416: check: data record [..., 466) overlaps txn boundary [..., 458)

As can be seen, in fs1 case, the error contains more details: [start, end) of
both mismatching transaction and data records.

In addition to fstest-like verbosity, add progress-mode, where % of total
completion is printed in a style similar to one used by `fs1 verify-index`.

The Go-based implementation is also faster even when data is on HDD. For
example on a 73GB database provided by @jerome[1] fsrefs.py takes ~15 minutes
to run and occupy ~70-100% of CPU. On the other hand `fs1 verify` takes ~7
minutes to run and occupy ~ 20-30% of CPU.

Tests pending.

[1] nexedi/zodbtools!19 (comment 129480)

Some dumpers might want to print something at the end of their dump.
We will need this functionality for Verify (see next patch).

And use that in the callers.

* master:
  go/zodb: Fix thinko in NoObjectError.Error

This error type is documented (see Loader) to be always created as
*NoObjectError.

-> Fix Error receiver accordingly.

* master:
  zodb: NewDB += option to set live cache control on created connections

WCFS needs this to run tests faster.

In general it is also a good idea to pass options to DB constructor, in
particular options that affect live cache size, or other properties, for
further created connections.

* master:
  go/*: Cosmetics
  go/xzodb: New package that provides zodb-related utilities shared by drivers  (draft)
  go/zodb/demo: New package that provides base+δ storages overlay
  go/zodb: Provide OpenDriver in addition to Open
  go/zodb/fs1: Report URL with file:// schema included
  go/zodb: Fix Open to detect lack of schema only by ":" instead of by "://"
  go/zodb: Fix IStorage implementation not to deadlock driver if watcher fails
  go/zodb/zeo: Fix it to provide "Close vs watchq" guaranty
  go/zodb/fs1: Fix it to provide "Close vs watchq" guaranty
  go/zodb: Require drivers to close watchq on Close
  go/internal/xtesting: Fix FatalIf to return a helper

For Load demo.Storage implementation is similar to DemoStorage in
ZODB/py with fixes "cherry-picked" from:

- https://github.com/zopefoundation/ZODB/issues/318
  (DemoStorage does not take whiteouts into account -> leading to data corruption)

- https://github.com/zopefoundation/ZODB/pull/323
  (loadAt + fix for the above issue)

For safety demo.Storage - contrary to DemoStorage/py - actually verifies
that for demo=base+δ δ comes strictly after base and that base remains
unchanged.

URI schema follows XRI Cross-references approach and is

	demo:(zurl_base)/(zurl_δ)

https://en.wikipedia.org/wiki/Extensible_Resource_Identifier provides
some related details and examples.

For ZODB/py corresponding pull-request for zodburi to support demo: URI
scheme has been made here: https://github.com/Pylons/zodburi/pull/29 .

Tests need:

- recent zodbtools with zodbrestore:

  https://lab.nexedi.com/nexedi/zodbtools/blob/129afa67/zodbtools/zodbrestore.py
  nexedi/zodbtools!19

- ZODB with support for DemoStorage.deleteObject
  https://github.com/zopefoundation/ZODB/pull/341

On Go side demo storage is needed for wendelin.core 2 because ERP5 uses
DemoStorage to run tests.

This is low-level API to open IStorageDriver instead of IStorage.
Demo storage will need this.

Maybe it would be a good idea to move drivers-related functionality into
separate package in the future.

In ZODB/go when there is no schema in zurl, open automatically prepends
file:// . However filename itself could contain ":" and so generally
speaking it is incorrect to return URL without file:// schema prepended
to file name.

Another reason to always use fully-constructed URLs with schema, is
interoperability with ZODB/py - there zodburi, when given zurl without
schema, does not make any assumption that it is of file:// kind and
rejects opening such URIs.

An URI schema is required to have ":" after it, but - even if frequently
used in practice - not //. We will soon introduce "demo:" URI scheme
that comes without //, so fix Open to detect schema presence just by ":"
and not to fixup "demo:..." url to "file://demo:..." automatically.

Before the patch if storage.watcher fails, storage.driver.Close is not
called, and so the driver will continue to send to .drvWatchq, but noone
is receiving from it.

a5dbb92b (go/zodb: Require drivers to close watchq on Close), provides
the guarantee that the driver will stop sending on drvWatchq right after
drv.Close call.