Support dedicated vector storage writes

[JingsongLi]

Summary

Add writer support for dedicated vector storage so tables configured with vector.file.format write vector columns into *.vector.<format> files instead of ordinary data files. This aligns append and data-evolution partial writes with the existing reader support for .vector.* files.

Changes

• Add vector.file.format and vector.target-file-size option accessors.
• Validate dedicated vector storage tables require data evolution, row tracking, and a normal anchor column.
• Split append writes so VECTOR columns are written to dedicated vector files while scalar/blob columns keep their existing layout.
• Split data-evolution partial writes by normal versus VECTOR columns so updates also produce dedicated vector files.
• Add focused coverage for .vector.parquet, .vector.vortex, option defaults, schema validation, and blob regression.

Testing

• cargo test -p paimon --lib vector
• cargo test -p paimon --lib blob_write
• cargo test -p paimon --features vortex --lib test_vector_write_uses_dedicated_vortex_file
• git diff --check

Notes

No migration is required. Existing inline VECTOR writes are unchanged when vector.file.format is not configured.

[leaves12138]

Looks good to me. I reviewed the updated dedicated vector storage writer changes, fixed the Clippy large-enum-variant issue by boxing the dedicated writer variant, and verified the relevant paths locally.

Validated locally:

• cargo fmt --check
• cargo clippy -p paimon --all-targets --features vortex -- -D warnings
• cargo test -p paimon --lib vector
• cargo test -p paimon --features vortex --lib test_vector_write_uses_dedicated_vortex_file

The full workspace Clippy check could not be reproduced locally because this environment lacks a dynamically embeddable Python for PyO3, but the CI blocker was isolated to the paimon crate and is fixed in the latest commit.