Fix tuple reconstruction of non-mappable dataclasses

[Linxiushen]

from_tuple() passes a zip object as the sole positional constructor argument. This works for mapping dataclasses, but with mappable_dataclass=False a multi-field dataclass raises TypeError, and a one-field dataclass silently stores the zip object instead of its value.

Construct from field-name keyword arguments so both supported dataclass modes receive the original values. Extend the tuple round-trip test across mapping/non-mapping, mutable/frozen and keyword-only constructors, and check that a restored single JAX-array field still works in a jitted function.

Validation on Python 3.12 / JAX 0.11.2 CPU:

• Unchanged source: 5 new cases fail, 4 mapping controls pass.
• python -m pytest chex/_src/dataclass_test.py -q --tb=short: 69 passed.
• The test.sh Flake8 selection and Google-configured Pylint on both changed files pass. Ruff reports the same two existing source diagnostics as the baseline, neither in changed code.

The full repository, docs, other Python versions and GPU tests were not run. This does not change pseudo-field or non-mapping init=False field behavior.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.