Dictionary page pruning for row filter predicates

## Summary

When a predicate targets a dictionary-encoded column, evaluate the predicate against the dictionary values before decoding any data pages. If no dictionary values match, the entire column chunk can be skipped (`AllFalse`). If all match, per-row evaluation can be skipped (`AllTrue`).

This adds an `evaluate_dictionary` method to the `ArrowPredicate` trait and integrates dictionary pruning into the row group filter evaluation pipeline.

Related PR: #9574

Start with the ArrowPredicate trait and the row group filter evaluation pipeline described in the issue; inspect how dictionary values and data pages are currently handled. The work is done when dictionary predicates produce AllFalse or AllTrue where applicable and avoid unnecessary per-row evaluation, with behavior covered by the relevant filter tests.

Written by the indexing model from the issue text.