Refactor MetaData class: reduce duplication and improve code quality

[pnorton-usgs]

Refactor MetaData class: reduce duplication and improve code quality

Summary

Major refactoring of metadata.py to reduce code duplication, modernize type hints, fix a latent bug, and improve readability.

Changes

Reduce duplication (largest change)

• Extract __filter_by_version() helper for version/deprecation filtering logic that was repeated in 5 of 6 parsing methods
• Extract __extract_common() helper for dimensions/modules/requires extraction repeated in 4 methods
• Extract __extract_valid_values() helper for valid values parsing repeated in 2 methods
• Remove ~40 lines of commented-out dead code
• Net reduction of ~90 lines while preserving identical behavior

Modernize type hints

• Add from __future__ import annotations
• Replace typing.Dict with built-in dict
• Replace Union[str, Version] with str | Version
• Remove unused typing imports

Replace implicit exception handling with explicit None checks

• Add __find_text() helper that safely returns element text or None
• Replace all elem.find(ev).text patterns wrapped in try/except AttributeError with explicit if text is None checks
• Makes control flow clearer and easier to reason about

Fix is_fixed bool parsing

• Replace bool() with a lambda that checks for '1', 'True', 'true'
• bool('0') and bool('False') both return True in Python, which would be incorrect if the XML ever uses those values
• Defensive fix (current XML only uses '1' for fixed dimensions)

Minor cleanup

• Remove unnecessary # type: ignore on xml.etree.ElementTree import
• Add __repr__ method for debugging
• Move outside_elem module-level dict to _OUTSIDE_ELEM class constant
• Move NEW_DTYPE and NEW_PARAM_DTYPE dicts to constants.py alongside other type-mapping dicts
• Add proper Sphinx-style docstring to __init__ and metadata property

Testing

All 290 tests pass. Metadata output verified structurally identical across versions 4, 5.2.1.1, 6.0.0, and 60.0.