refactor: use REE field name constants across crates

[akashchamp]

Which issue does this PR close?

• Closes Wire REE field name constants to more usages in arrow-schema #11035.

Rationale for this change

Field::REE_RUN_ENDS_FIELD_DEFAULT_NAME and Field::REE_VALUES_FIELD_DEFAULT_NAME already exist in arrow-schema, but the default run_ends / values child names were still spelled out as string literals across the workspace. Using the constants everywhere keeps the default names in one place, in the same way #10517 did for the map field names.

What changes are included in this PR?

Replace every default run-end encoded child field name literal with the shared constants, keeping behaviour identical:

• arrow-schema: display, parse, ffi, fields and canonical extension code and tests
• arrow-array: RunArray::try_new (the default names every RunArray gets) and new_null_array tests
• arrow-data: run transform and validation tests
• arrow-cast, arrow-ord, arrow-row, arrow-select (no REE literals there), arrow-ipc, arrow-json, arrow-avro (codec, schema and writer round-trip test), arrow-integration-test (JSON runendencoded placeholder type), arrow (data_gen and the cast benchmarks), parquet (arrow_writer bench) and parquet-variant-compute tests

Intentional custom-name cases (for example the named_values display/parse tests) and unrelated "values" field names (dictionary values in arrow-flight, map value fallback in arrow-avro, plain column names in arrow-ipc tests) are left as literals on purpose.

Are these changes tested?

Covered by the existing tests; no behaviour changes. Run locally:

• cargo fmt --all -- --check
• cargo test --lib for arrow-data, arrow-array, arrow-ord, arrow-row, arrow-ipc, arrow-json, arrow-cast, arrow-avro, arrow-integration-test, arrow (with test_utils) and parquet-variant-compute
• cargo check for the touched arrow, arrow-json and parquet benchmarks
• cargo clippy --all-targets --all-features -- -D warnings on the touched crates

Are there any user-facing changes?

None.

AI assistance was used to identify and apply the mechanical substitutions across crates. I reviewed every changed construction and confirmed that custom-name cases and unrelated "values" field names remain explicit literals.

[Rich-T-kid]

seems straightforward, thank you @akashchamp

[Jefffrey]

i think we should aim to rewire to more crates than just arrow-schema, like how maps was done

• formalize the default map field names to match default arrow spec #10517

[akashchamp]

Good call, thanks for the pointer to #10517. I have now rewired the rest of the workspace the same way: arrow-array (RunArray::try_new), arrow-data, arrow-cast, arrow-ord, arrow-row, arrow-ipc, arrow-json, arrow-avro, arrow-integration-test, arrow, parquet and parquet-variant-compute all use Field::REE_RUN_ENDS_FIELD_DEFAULT_NAME / Field::REE_VALUES_FIELD_DEFAULT_NAME, so no default run_ends/values literals remain outside the constant definitions. arrow-select and arrow-flight had no REE name literals to change, and the remaining "values" literals are dictionary, map or ordinary column fields, not REE children, so I left them as they are.

[Rich-T-kid]

this looks good, nice suggestion Jefffrey