Validate and reconstruct namedtuple fields in ExactSequence

[xingzhao-ji]

ExactSequence accepts tuple inputs but reconstructs them with type(value)(generator). For named tuples, this either raises TypeError (zero or multiple fields) or silently stores the generator in a one-field record without validating its contents.

Use the same named-tuple detection and positional reconstruction already used by ordinary tuple schemas. Keep ordinary list/tuple reconstruction and validation error handling unchanged.

Five regression cases cover zero-, one-, and two-field named tuples, coercion, preserving the record type and original input, and invalid fields with default/custom messages inside a dictionary schema. The API docstring and changelog describe the fix.

Validation:

• All five new cases fail on the original implementation and pass with the fix.
• Configured test selection: 187 passed on Python 3.12.14 and 3.13.5.
• Explicitly including both test modules (the configured filename filter excludes test_dataclasses.py): 205 passed on each Python version.
• Independent review ran six ExactSequence tests and both test modules successfully.
• Mypy 1.16.0 passes all 10 source files; git diff --check passes.
• An additional module/README doctest run has three failures, reproduced unchanged on pristine HEAD. Ruff checks likewise report only existing E721 comparisons and legacy Unicode-prefix formatting differences.

Black, isort, flake8, and the full supported Python/OS matrix were not run in the existing environment. No claim of completed upstream CI is made.

This extends the existing named-tuple handling from #231 to ExactSequence; it does not change unordered-sequence matching.