[terbed]

Fixes #62

Forward datafusion.execution.parquet.enable_page_index to Iceberg's
scan builder. Iceberg scans now follow the session setting, including
DataFusion's enabled default.

Add regression coverage for reduced Parquet reads with identical query
results, explicit disabling, files without page indexes, and nullable
decimal columns with projection. Tests cover both static and
catalog-backed table providers.

Update the Iceberg dependency to include apache/iceberg-rust#3314, which prevents incorrect row elimination for decimal predicates when page-index pruning is enabled. Adjust existing tests for the updated dependency’s API changes.

[terbed]

Thanks @gabotechs for taking the time to review and comment on this PR.

• I've removed the version bump
• Removed tests proving page index pruning reduces byte reads
• Keeping correctness tests using the taxi dataset with and without page indexing enabled

This approach significantly simplified the pr, the main complexity is proving page index pruning's effect. That could be a separate focused PR if such test would be important/necessary.

f95e1a6

Would you recommend further changes? Or could this PR in this form be merged?

[gabotechs]

👍 All good here! thanks @terbed. Will merge it as soon as CI passes.