fix: handle bytes values in string column statistics from Parquet

[GabrielAmazonas]

Rationale for this change

When using add_files() to import Parquet files written by DuckDB into PyIceberg tables, the operation fails with AttributeError: 'bytes' object has no attribute 'encode'.

This occurs because the Parquet specification defines column statistics (min_value, max_value) as binary data:

struct Statistics {
  5: optional binary max_value;
  6: optional binary min_value;
}

This change is a follow-up improvement to #1354, which fixed handling of missing column statistics. This PR addresses the case where statistics are present but returned as bytes instead of str.

This change is indirectly related to broader DuckDB-PyIceberg interoperability challenges documented in duckdb/duckdb#12958

When PyArrow reads these statistics from Parquet files, it may return them as Python bytes objects rather than decoded str values, which is valid per the Parquet spec. However, PyIceberg's StatsAggregator only expected string statistics to be str objects, causing failures when processing files from writers like DuckDB that expose this binary representation.

This PR fixes the issue by adding proper handling for bytes values in string column statistics:

1. StatsAggregator.min_as_bytes(): Decode bytes to UTF-8 before truncation and serialization
2. StatsAggregator.max_as_bytes(): Decode bytes to UTF-8 before processing (previously raised ValueError)
3. to_bytes() for StringType: Add defensive isinstance check as a safety fallback
4. Add comprehensive unit tests for both StatsAggregator and to_bytes

This improves interoperability with DuckDB and other Parquet writers that expose statistics in their binary form.

Are these changes tested?

Yes, this PR includes unit tests that verify:

• StatsAggregator.min_as_bytes() correctly handles bytes values for string columns
• StatsAggregator.max_as_bytes() correctly handles bytes values for string columns
• to_bytes() for StringType correctly handles both str and bytes inputs

The tests ensure the fix works correctly while maintaining backward compatibility with existing str-based statistics.

Are there any user-facing changes?

Yes - This is a bug fix that enables users to successfully use add_files() with Parquet files written by DuckDB and potentially other writers that return statistics as bytes.

Before: add_files() would fail with AttributeError: 'bytes' object has no attribute 'encode' when processing DuckDB-written Parquet files.

After: add_files() correctly processes Parquet files regardless of whether statistics are returned as str or bytes.

This change is backward compatible - existing workflows with Parquet files that return str statistics will continue to work as before.