fix(parquet): avoid panic on ColumnIndex length mismatch

[pchintar]

Which issue does this PR close?

• Closes ColumnIndex length mismatch can cause panic during decoding in Parquet #9832 .

Rationale for this change

In parquet/src/file/page_index/column_index.rs, ColumnIndex decoding assumes that page-aligned arrays (null_pages, min_values, max_values, and optional arrays) have matching lengths, but this is not validated.

As a result, malformed metadata can trigger an out-of-bounds panic during decoding instead of returning a ParquetError. Since parquet files are external input, this should be handled safely.

What changes are included in this PR?

• Added validation in:

  • PrimitiveColumnIndex::try_new
  • ByteArrayColumnIndex::try_new

• Ensures:

  • min_values.len() == null_pages.len()
  • max_values.len() == null_pages.len()
  • optional arrays (null_counts, histograms) are consistent with page count

• Returns ParquetError on mismatch instead of panicking

Are these changes tested?

Yes.

Added a unit test:

• test_column_index_rejects_mismatched_min_max_lengths

This constructs a ColumnIndex with mismatched lengths and verifies that decoding returns an error instead of panicking.

Are there any user-facing changes?

No.