validate tagged objects using schema for tag

[braingram]

Passing a Tagged object to a validator with a custom schema fails to validate the Tagged object using the schema associated with the tag.

This bug is impacting roman_datamodels (which uses this pattern for assignment validation). Without this PR the following will not raise a ValidationError:

import roman_datamodels.maker_utils

m = roman_datamodels.datamodels.ImageModel(roman_datamodels.maker_utils.mk_level2_image())
inst = m.meta.instrument
inst['name'] = 'a'
m.meta.instrument = inst

With this PR a ValidationError is raised ('a' is not one of ['WFI']).

This PR adds a test for that issue and fixes it by shuffling around a few items in the validator (so that tags trigger schema lookups even if a custom schema was provided to the validator).

This shuffling had the side effect of changing the validation error for one test (test_history) to what looks like a more helpful message.

This PR includes some changes from #1729 which should be merged first.

Running roman regtests as this will change how validation on assignment behaves: https://plwishmaster.stsci.edu:8081/job/RT/job/Roman-Developers-Pull-Requests/605/

Checklist:

• pre-commit checks ran successfully
• tests ran successfully
• for a public change, a changelog entry was added
• for a public change, documentation was updated
• for any new features, unit tests were added

[braingram]

The roman_datamodels downstream job is now failing with this PR with a 143 exit code.

Running these locally they pass but use a considerable amount of RAM (>8GB).

Running the same tests with the released asdf takes ~600MB.

This was only seen in recent runs so something has changed in roman_datamodels that now conflicts with the changes in this PR. I'm going to convert this to draft until the cause of the memory issue can be sorted.