Optimize repeated Map lookups with lazy index - #16

Motivation

• Repeated WDL Map key lookups were doing a linear scan of Value.Map.value on every access, which is inefficient for common cases with many lookups.
• Make a small, low-risk optimization that reuses a Python dict index after the first lookup to improve performance without large refactors.

Description

• Add cached index fields to WDL.Value.Map: _index and _index_failed, and initialize them in __init__.
• Implement WDL.Value.Map.get(key) which lazily builds _index on first lookup, uses _index for subsequent lookups, and falls back to the original linear scan when keys are unhashable in Python.
• Provide _index_key() helper that derives the Python dict key from a Value.Base (currently returns key.value).
• Switch stdlib map access in _At._call_eager to use Value.Map.get() so the optimization applies to map[key] evaluations.

Testing

• Ran formatting and static checks: ruff format WDL/Value.py WDL/StdLib.py and ruff check WDL/Value.py WDL/StdLib.py, which passed.
• Ran type checks with mypy WDL/Value.py WDL/StdLib.py, which passed.
• Ran targeted unit tests with pytest -q tests/test_0eval.py -k map, which passed (1 passed, 22 deselected).

Codex Task