Collect ChgrpValidation exceptions

[jamoore]

When the first validation error (foreign key constraint) is encountered during graph processing, an exception is immediately thrown and the ID of the object causing the exception is returned in the constraints field of the omero::cmd::ChgrpERR object. Clients would benefit from having all the constraints rather than just the first such constraint.

Currently there is nothing in place to allow this type of detection. A possible workaround may be to perform a query (see #10782) before executing the chgrp.

See:

• #10776
• the testTwoFilesetsChgrp* methods of ​PR 1139
• see tests disabled in ​PR 1256

[jburel]

Moved from sprint FS demo 4.2

[jamoore]

With Mark's work on #10782 / #10890, this should no longer be of significant impact. We may want to re-write the tests as a part of that work to not look for a constraint, but rather to perform the query. The constraint checks could be left in / turned back on as a double validation, once this ticket is handled.

[jmoore <josh@…>]

(In [92b69fe02b0a709f5fdd5aa34974ac4132cdd833/ome.git] on branch develop) Add finish() phase to IRequest (See #11045, #11054, etc)

Now that Fileset handling is happening after containers and
images, it's necessary to have a phase hook which checks that
all filesets have been deleted along with the related images.

This amounts to a trigger pre-commit, but that's not provided
by postgresql. Instead, we collect all of the fileset IDs in
the DeleteStep? instances and then check that they have been
removed afterwards. The primary drawback is that this no longer
provides a GraphConstraintERR but rather just a regular ERR.

See #10846

Note: if this requirement is ever removed, the entire finish()
phase could be rolled back. Alternatively, this could replace
the ChgrpValidation? steps.

[jamoore]

This has been invalidated by the new Chgrp2 implementation.