Fix #653: Handle UTF-8 BOM transparently in all file loading paths

[javihern98]

Summary

Use Python's utf-8-sig encoding in all user-facing file read paths to transparently strip UTF-8 BOM when present. Additionally, strip BOM from DataFrame column names in _validate_pandas to handle cases where users pass DataFrames read from BOM-encoded CSVs without utf-8-sig.

Previously, BOM files caused misleading errors (corrupted CSV column names, JSON parse errors, ANTLR syntax errors) with no indication of the actual encoding issue.

Fixes #653

Checklist

• Code quality checks pass (ruff format, ruff check, mypy)
• Tests pass (pytest)
• Documentation updated (if applicable)

Impact / Risk

• Breaking changes? No — utf-8-sig is a strict superset of utf-8 for reading. Files without BOM behave identically. The column name stripping only affects the \ufeff character at position 0.
• Data/SDMX compatibility concerns? None. Only read paths are changed; write paths remain utf-8 (no BOM added to output).
• Notes for release/changelog? Files saved with UTF-8 BOM (common from Windows editors) now load correctly instead of producing misleading errors. DataFrames with BOM-contaminated column names are also handled.

Notes

• Removed unused error code 0-1-2-5 which was defined but never raised.
• utf-8-sig only inspects the first 3 bytes — no performance impact.
• Write paths (to_csv, scalar output) are intentionally left unchanged since utf-8-sig would add a BOM to output files.
• The BOM stripping in _validate_pandas covers all DataFrame entry points: direct DataFrames via run(), URL-fetched DataFrames, and file-loaded CSVs.