Also update README accordingly.
Its becasue now erp5-site is being created automatically using software-type
default and there is no create-erp5-site software-type anymore.

In the particular case of PATH, '' is the same as '.',
which is a bad idea from a security point of view.

instance-erp5.cfg.in:
  quoted values should not be part of the configuration parameters
    https://lab.nexedi.com/nexedi/slapos.toolbox/blob/beb86eab2dee7178aa6e4795da0f56bb6cacaea1/slapos/apachedex.py#L62
  modify default apachedex parameters
    --erp5-base +erp5 .*/VirtualHostRoot/erp5(/|\?|$)
    --base +other /
instance-balancer.cfg.in:
  option `--configuration` should be used instead of `--config`
    https://lab.nexedi.com/nexedi/slapos.toolbox/blob/beb86eab2dee7178aa6e4795da0f56bb6cacaea1/slapos/apachedex.py#L76

instance-mariadb.cfg.in:
  rename slowquerydex folder to slowquery_digest
mysql-querydigest.sh.in:
  exit 1 on error
  add date to the output file name of pt-query-digest to fit promise behavior
    https://lab.nexedi.com/nexedi/slapos.toolbox/blob/beb86eab2dee7178aa6e4795da0f56bb6cacaea1/slapos/promise/check_slow_queries_digest_result/__init__.py#L29
  review sh arguments handling
  end message: "ok" -> "Report generated successfully." or ..."failed with code $?"

https://cgit.freedesktop.org/fontconfig/commit/?id=1ab5258f7c2abfafcd63a760ca08bf93591912da

fontconfig > 2.12.1 now requires gperf.

This reverts commit 38123b86.

fontconfig was not building any more on many machines (most testnode machines). Revert
until investigation is done.

https://patchwork.openembedded.org/patch/134994/

This fixes part of resiliency tests

/reviewed-on nexedi/slapos!262

( because  allow-picked-versions require all versions to be pinned )

Since erp5 commit "[taskdistribution] Drop needless TaskDistributionTool"
the tool is not available anymore and only TaskDistributor shall be used

It provides fix to call correct endpoints (Distributor instead of a Tool).

/reviewed-on nexedi/slapos!260

See also https://bugs.debian.org/882329

Issue was:
- user project_user was created
- project_user create stored routines
- mysql data is dumped (including list of users)
- mysql data is restored, but stored routines have the information
  DEFINER that must match an existing user. This does not
  work if there is no flush privileges instructions

So make sure to insert flush privileges while dumping

This reproduces what happens on a Prod system, where
TokuDB turned out to use much more CPU than FileStorage.