fix(serde): align DataFile JSON with Java core

[kamcheungting-db]

Summary

Make core DataFile JSON compatible with Java Iceberg.

This PR only changes DataFile JSON behavior. Core FileScanTask serde is added separately in #939.

Changes with examples

Partition data

Continue accepting positional arrays:

"partition": [7]

Also accept Java field-ID objects, which do not depend on field order:

"partition": {"1000": 7}

Binary metadata

Write binary values as Java-compatible hexadecimal strings:

"key-metadata": "0A0B"
"lower-bounds": {"keys": [1], "values": ["01000000"]}

Previously these values were written as JSON byte arrays.

Content names

Accept both current lowercase values and older Java enum-style values:

"content": "position-deletes"
"content": "POSITION_DELETES"

Integers

Reject values that are not actual, in-range JSON integers. For example:

"record-count": 1.5

Previously generic JSON conversion could coerce this to 1.

The parser also accepts a missing partition, accepts null metric maps, rejects duplicate metric keys, and validates partition value types during serialization.

Stack

This is 2 of 4 and depends on #974:

1. Move DataFile serde into core (refactor(serde): move DataFile JSON serde into core #974)
2. This PR: Java-compatible DataFile JSON
3. Add core FileScanTask serde (feat(serde): add core FileScanTask JSON support #939)
4. Validate schema and partition IDs strictly (fix(serde): validate schema and partition IDs as JSON integers #940)

Each PR targets main; later PRs contain the commits from earlier PRs.

Testing

• DataFile focused tests: 23 passed across core and REST.
• Full stack: 142 JSON serde tests passed.
• Full stack: 332 REST tests passed; 1 existing test skipped.