fix: honor explicit temperature when config= or model= is passed to extract()

[fsaudm]

Description

extract() accepts a top-level temperature argument documented as "Set to 0.0
for deterministic output", but the value is only merged into provider kwargs in
the model_id branch of extraction.py. When a caller passes a pre-built
config= (factory.ModelConfig) or model= (a language model instance), the
argument is silently ignored: the request carries no temperature parameter at
all (the OpenAI provider omits it when the value is None), so generation runs
at the server default temperature.

We found this with a batch-invariant vLLM deployment: raw API probes at
temperature 0 were byte-identical across runs, while extractions through
langextract differed run to run. A wire capture showed the request parameters
were {"model", "n", "response_format"} with no temperature despite
temperature=0.0 being passed.

This PR:

• config= branch: merges an explicit temperature into the config via
  dataclasses.replace, so the caller's ModelConfig is never mutated. An
  explicit kwarg wins over a config-supplied value; with no explicit kwarg the
  config value is untouched.
• model= branch: emits a UserWarning instead of silently ignoring the
  argument (a constructed model has no provider-agnostic temperature setter, and
  silently mutating caller-configured state seemed worse). This mirrors the
  existing use_schema_constraints warning in the same branch.
• Updates the temperature, config, and model docstrings with the
  precedence rule.

Fixes #529

Choose one: Bug fix

How Has This Been Tested?

Added tests/extract_precedence_test.py with four tests covering the
precedence rules: explicit kwarg wins over a config-supplied value, a
config-supplied value survives when no kwarg is passed, the caller's
ModelConfig is not mutated, and the model= branch warns. Ran the new file
and the full suite:

$ pytest tests/extract_precedence_test.py -v
$ pytest tests

All pass. The original symptom was also re-verified end to end against our
batch-invariant vLLM deployment: with the patch, extractions at
temperature=0.0 are byte-identical across runs, matching the raw API probes.

Checklist:

• I have read and acknowledged Google's Open Source
  Code of conduct.
• I have read the
  Contributing
  page, and I either signed the Google
  Individual CLA
  or am covered by my company's
  Corporate CLA.
• I have discussed my proposed solution with code owners in the linked
  issue(s) and we have agreed upon the general approach.
• I have made any needed documentation changes, or noted in the linked
  issue(s) that documentation elsewhere needs updating.
• I have added tests, or I have ensured existing tests cover the changes
• I have followed
  Google's Python Style Guide
  and ran pylint over the affected code.

[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.

[fsaudm]

@googlebot I signed it!

[github-actions]

⚠️ Branch Update Required

Your branch is 1 commits behind main. Please update your branch to ensure CI checks run with the latest code:

git fetch origin main
git merge origin/main
git push

Note: Enable "Allow edits by maintainers" to allow automatic updates.

[github-actions]

⚠️ Branch Update Required

Your branch is 4 commits behind main. Please update your branch to ensure CI checks run with the latest code:

git fetch origin main
git merge origin/main
git push

Note: Enable "Allow edits by maintainers" to allow automatic updates.

[github-actions]

⚠️ Branch Update Required

Your branch is 7 commits behind main. Please update your branch to ensure CI checks run with the latest code:

git fetch origin main
git merge origin/main
git push

Note: Enable "Allow edits by maintainers" to allow automatic updates.