Support all xarray coordinate types in ScalarBound

[alxmrs]

Problem

The ScalarBound enum in src/lib.rs (marked TODO at line ~78) only handles three coordinate types:

• Int64 (integer coordinates)
• Float64 (lat/lon coordinates)
• TimestampNanos (datetime64[ns] coordinates)

xarray supports a much wider range of coordinate types. Unrecognized types silently fall through with no partition pruning.

Missing types

• UInt8, UInt16, UInt32, UInt64 (unsigned integers)
• Int8, Int16, Int32 (smaller integer widths)
• Float32 (single-precision float)
• Timedelta / duration types
• String / categorical coordinates (e.g. station names, model names)
• Boolean coordinates

Work required

1. Extend ScalarBound with variants for the missing types
2. Update compare_to_scalar to handle each new variant against matching DataFusion ScalarValue types
3. Update python_to_scalar_bound to extract these types from Python objects
4. Add partition_metadata() conversion in df.py for new types
5. Add tests covering each new coordinate type

Acceptance criteria

• Partition pruning works correctly for each supported xarray coordinate dtype
• Unsupported types degrade gracefully (no pruning, no error)