[Bug] Bitmap file index drops rows for <> / NOT IN on TIMESTAMP(7..9) columns

[thswlsqls]

Search before asking

• I searched in the issues and found nothing similar.

Paimon version
master @ bbae07d

Compute Engine
Any engine writing TIMESTAMP(9) / TIMESTAMP_LTZ(9) columns (Flink, Java API); Spark's microsecond default is unaffected.

Minimal reproduce step

1. Create a table with f1 TIMESTAMP(9) and file-index.bitmap.columns = f1.
2. Insert two rows whose f1 differ only below a microsecond, e.g. ...00.000123000 and ...00.000123400.
3. SELECT * FROM t WHERE f1 <> TIMESTAMP '...00.000123000'.

Expected 1 row, got 0. BitmapFileIndex.getValueMapper() keys TIMESTAMP / TIMESTAMP_LTZ through Timestamp.toMicros(), so both rows share one key; Reader.visitNotIn flips that bitmap and RawFileSplitRead applies it as an exact row selection, dropping the whole bucket.

What doesn't meet your expectations?
<> / NOT IN on an indexed TIMESTAMP(7..9) column must return the same rows as an unindexed scan. #9654 fixed this for BSI (REMAIN for value predicates, exact IS NULL / IS NOT NULL); bitmap has used the same mapper since #3775; range-bitmap rejects precision > 6 at build time.

Anything else?
N/A

Are you willing to submit a PR?

• I'm willing to submit a PR!