when stopping an erp5 cluster instance, other sub instances like mariadb, zodb, etc. will stay
started. Only the root instance will be stopped. This is because the root instance doesn't send
his state to children.

Add state in default requested now allow sub instances to be requested with parent state.

Reported by Julien

The feature is postponed and noted in TODO.

The `gzip` declaration uses defaults, which seems reasonable.

/reviewed-on nexedi/slapos!352

The md5sum was not updated, so update it ASAP.

In case of error during processing SlapOS software or instance expose
last 100 lines of the message, which is enough to identify the problem and
does not clutter the test master.

@vincentB @kazuhiko 

/reviewed-on !351

/reviewed-on nexedi/slapos!350

Commit c6c919dc (Remove all occurrences of the obsolete
'parameters-extra' option) removed

	parameters-extra = true

because after 5c980caf (Drop the 'parameters-extra' option and always
forward extra parameters) "$@" is always implicitly added to argv.

Clarify the comment that was explaining what `parameters-extra = true`
was doing, since it was not updated during parameters-extra removal.

Features:

 * amend configuration with comments
 * drop obsolete comments from Apache copy
 * remove not needed whitespaces
 * use indentation for conditionals in Jinja2

Where applicable simplify Jinja2 templates by removing Caddy-TODO
and fill or update TODO.rst.

Therte is no need to control whitespace adding by Jijna2 and dropping it
simplifies the templates.

It will allow to take better control over generated configuration files.

The goal is to drop slapos.cookbook:erp5.test

No need for a recipe to do a bit of url parsing and generating 2 wrappers.
Also, drop unused & empty "host-common" section.

- stack/erp5: Add path to all binary links in PATH
- stack/erp5: Add single quotation marks to not break parameters in many lines
- stack/erp5: Fix path to avoid one section to overwrite the file from another section
- stack/erp5: pass cloudooo host instead of the IP address. The IP address and cloudooo hostname are well in $SLAPPART/etc/hosts
- stack/erp5: Use use-ipv6 instance parameter to control where caucased listens when locally deployed.
  Drop incomplete 6tunnel usage on balancer toward zopes.

Tests are somehow happy test_result_module/20180626-62AD32BB test_result_module/20180626-70AC16FA

If no objections I would like to merge tomorrow but feel free to press merge for me :)

The goal is rebase nexedi/slapos!318 to fix the conflict and merge after tomorrow.

/cc @rafael @vpelletier 

/reviewed-on nexedi/slapos!347

While reading templates/cached-virtualhost.conf.in it seems, that
SSL-enabled host shall be exposed, but this is not true -- it is
connecting to SSL backend, but listening on non-SSL.

In order to stabilise this add assertsion to test_enable_cache
which access cached ports directly and show its functionality, thus
resolving added TODO.

Whenever there is a problem while building software or instantiation,
emit verbose information for further analysis.

apache-frontend supports correctly ssl_proxy_ca_crt for ssl-proxy-verify,
thus show it in test, still keeping expected 501 reply for caddy-frontend.

The schema shall be https, not http. It got undetected, as caddy-frontend
dies really fast with 501, while apache-frontend is able to communicate
with the backend.