When indexing an entity type, don't generate a single
and possibly huge rset with all entities of that type.
Instead, fetch entities chunk by chunk and free caches
between each round.

closes #17090555

ESTransactionQueue batches all ES operations on postcommit.

Client cubes could override the default logic to add, log,
ditch… any other operation.

mocked parameter list was inverted

and it seems redundant with `allow_all_hooks_but`

Default is to use 'eid' but one may want to use something
else (stable id, ark, etc.)

closes #17079908

* get rid of CUSTOM_ATTRIBUTES. Replace it with IFullTextIndexSerializable adapter `custom_indexable_attributes`
  attribute;
* FullTextIndexSerializable.serialize method : default complete value is now True*


closes #17079905

CubicWebConfig.config_for might accept extra parameters such
as "debugmode"

--debug and --dry-run should be used as flag-only options.
They don't expect a "yes" / "no" value.

Also, remove tests on stdout in debug mode because it's
fragile and messes up with logging handling in cubicweb.

It's not used (and therefore not real-life-tested) anywhere
and current implementation is known to be partial and a bit
hard to maintain.

(closes #17071989)

closes #17052625

FTI update is now done at commit time in an operation,
we need to call commit() to trigger it

[test] only check composed search doesn't crash, the query is too unstable to maintain tests for now

Also : ccplugin - generator serves all actions, not split up by type
       es.py - add CUSTOM_ATTRIBUTES mechanism