[schema] properly raise BadSchemaDefinition when some expression mains variables may not be guessed

[CWEP002] properly handle serialization of computed relations We now: * have CWComputedRelation in the bootstrap schema to store computed relations * properly serialize/deserialize it * test first if the database has been migrated and contains the related table Related to #3546717 [jcr: adjust unittest_querier to pass with the added entity type]

[CWEP002] Plug the computed relation rewriter in the querier Related to #3546717.

[CWEP002] refactor rql read security checking Split 'check_read_perms' into 'check_relations_perms' which checks relations 'read' permissions and 'get_local_checks' which build dictionary of local security checks (rql expression) for variables. This allows to check relations 'read' permissions earlier in the process and so to prepare insertion of the rql rewriter: we want to check permissions of the computed relation, not permissions of relations introduced by the associated rule, to conform to the CWEP. Related to #3546717

[CWEP002] introduce RQLRelationRewriter Refactor existing RQLRewriter for later reuse of rewriting relation as specified by CWEP002. Work is different because we simply want to replace a relation by another rql snippet and we don't have to bother with EXISTS, subqueries and all. This rewriter is not yet plugged into the querier. Depends on yams 0.40 API. Related to #3546717

[CWEP002] Add schema finalization checks for computed relations (rules) Related to #3546717

[schema] add utility function to build a CubicWebSchema from a namespace Will be useful for testing computed attributes and relations.

Fix test migration crash waiting to happen due to inferred relations Avoid crashing when a relation definition which is already known (due to being inferred) gets explicitly added. As explained in a comment, this would deserve more thinking, but at least this fixes a test failure with computed relations.

fix typo in syncschema hooks

[test] Make test_undo_api less random The order in which hooks are run is not predictable if they have the same 'order' attribute, which is the case for SetOwnershipHook and SetInitialStateHook. So don't assume in_state will be set before created_by.