persist: Plumb relevant arrow::DataTypes to compaction

[ParkMyCar]

This PR plumbs arrow::datatypes::DataType from Persist's schema registry down into compaction.

Previously we changed compaction to use the most recent schema of any since run (#30382). This would pass the relevant K::Schema and V::Schemas down into compaction, but then map to an arrow::DataType using the current code.

The problem we're running into when trying to stabilize schema evolution in Persist (#30205) is that during 0dt upgrades the new version of Materialize ends up getting some compaction work assigned to it, and writes a batch of data using more nullable types. The old version of Materialize sees the new schema, but maps the arrow::DataType to less nullable types because we haven't yet changed how RelationDesc maps to an arrow::DataType.

After this PR the old version of Materialize should use the more up-to-date version of the arrow::DataType mapping and succeed.

Note: If this change is undesirable, we can remove it after a single version upgrade.

Motivation

Fixes an issue discovered in: #30205

Checklist

• This PR has adequate test coverage / QA involvement has been duly considered. (trigger-ci for additional test/nightly runs)
• This PR has an associated up-to-date design doc, is a design doc (template), or is sufficiently small to not require a design.
• If this PR evolves an existing $T ⇔ Proto$T mapping (possibly in a backwards-incompatible way), then it is tagged with a T-proto label.
• If this PR will require changes to cloud orchestration or tests, there is a companion cloud PR to account for those changes that is tagged with the release-blocker label (example).
• If this PR includes major user-facing behavior changes, I have pinged the relevant PM to schedule a changelog post.

[bkirwi]

Trying to think through what happens when a recent version ends up calling get_or_make_structured on an old batch without structured data, using the old schema with the old datatype... IIUC it will use the current encoder logic to generate a nullable version of that data, then try and datatype-migrate it to the old datatype and fail?

[ParkMyCar]

No longer needed!