Include metadata in numpy/polars cache fingerprints to prevent collisions

[skrawcz]

Summary

Fix hash collisions in the caching subsystem's fingerprinting for numpy arrays and polars DataFrames.

Problem

• hash_numpy_array used only obj.tobytes(), which discards shape and dtype. Arrays with identical raw bytes but different shapes (e.g., shape=(6,) vs shape=(2,3)) or different dtypes (e.g., float32(1.0) vs int32(1065353216)) produced identical cache keys.

• hash_polars_dataframe used only obj.hash_rows(), which discards column names. DataFrames with identical cell values but different schemas produced identical cache keys.

Both could cause the cache to silently return incorrect results from a previous execution.

Fix

• hash_numpy_array: prepend f"{obj.shape}:{obj.dtype}" to the bytes before hashing
• hash_polars_dataframe: include column names and dtypes (schema) alongside row hashes

Backwards compatibility

This changes hash output for numpy arrays and polars DataFrames. Existing caches will miss (different hash = recomputation), not produce incorrect results. Users will see a one-time recomputation after upgrading but no manual cache clearing is needed.

Tests

Added tests verifying:

• Different shapes produce different hashes
• Different dtypes with same bit pattern produce different hashes
• Different column names produce different hashes
• Identical data still produces identical hashes

Reported-by: Dem0