[repository api] definitly kill usage of word 'pool' to refer to connections set used by a session Also document session's data storage and some other internals. Hopefuly things will get clearer. Closes #1684860: vocabulary confusion in repository code: notion of 'pool'

backport stable

[rql2sql] fix in 7394:d11d88bd08e8 was actually not needed

[rset] close #1683703: rset.get_entity crash on rql query with subquery and aggregat

[rql2sql] fix generated sql for eid comparison. closes #1638695

fix some deprecation warnings

[querier test] test groupby function call / column number

backport stable

[plot views] don't paginate

backport oldstable

[url rewriting] fix nasty bug in build_rset inner function: the same dictionary is used by all queries, hence triggering potentially important bugs on concurrent requests (this dictionary goes back and forth the repository)

[hooks selection optimization] prune hooks when multiple entities are concerned by a hm.call_hooks() (closes: #1672022) the idea is to make a first pass over all the hooks in the registry and to mark put some of them in a disabled list. The disabled hooks are the one which: * are disabled at the session level * have a match_rtype or an is_instance selector which does not match the rtype / etype of the relations / entities for which we are calling the hooks. This works because the repository calls the hooks grouped by rtype or by etype when using the entities or eids_to_from keyword arguments Only hooks with a simple selector or an AndSelector of simple selectors (is_instance and match_rtype) are considered for disabling.

[repo test] datafeed parser's process method should now take raise_on_error argument

[datafeed] propagate raise_on_error to parser's process method