Share Python legacy model helpers across generated models

[Abhinav0905]

What changed

• Generate the Python legacy model conversion helpers once in a private _legacy_model_helpers.py package module for full client generation.
• Import the shared helpers from generic, anyOf, and oneOf models and from ApiClient.
• Keep model-only generation self-contained by falling back to the existing inline helpers when supporting files are not generated.
• Add focused generator assertions and regenerate the legacy model dictionaries sample.

Why

compatibleWithPythonLegacy currently expands the same helper partial into every generated model and ApiClient. In kubernetes-client/python#2677, this accounts for roughly 1,538 repeated model copies and about 3.4 MB of duplicated source across the synchronous and asynchronous clients.

This fixes the duplication at its source without adding a generator option or changing the generated public API. The small checked-in sample drops from 151,102 to 140,230 Python source bytes (7.20%) and from seven copies of _get_openapi_to_dict to one.

Closes #24729.

Downstream follow-up: kubernetes-client/python#2677

Validation

• mvn -ntp -pl modules/openapi-generator -am -Dtest=PythonClientCodegenTest#testLegacyModelToDictRendering+testLegacyModelToDictSupportsModelOnlyGeneration -Dsurefire.failIfNoSpecifiedTests=false test — 2 passed
• mvn -ntp -pl modules/openapi-generator-cli -am package -DskipTests=true -Dmaven.javadoc.skip=true -Djacoco.skip=true — build succeeded
• ./bin/generate-samples.sh bin/configs/python-legacy-model-dictionaries.yaml
• python -m pytest -q tests in the regenerated sample — 11 passed
• python -m flake8 legacy_model_dict_client/_legacy_model_helpers.py
• git diff --check

PR checklist

• Read the contribution guidelines.
• Run the full-project build, regenerate every sample, and export all generator docs. The focused Python sample and relevant modules were built and tested locally; the complete matrix is left to CI while this PR is in draft.
• Python technical committee: @cbornet @tomplus @krjakbrjak @fa0311

---

Summary by cubic

Shares Python legacy model conversion helpers across generated models to remove duplication without changing the public API. Previously each model and ApiClient inlined the helpers; now full client generation emits a single private _legacy_model_helpers.py that models and ApiClient import, with a model-only fallback to inline helpers.

• Generate _legacy_model_helpers.py when legacy compatibility is enabled and supporting files are generated or explicitly requested via CodegenConstants.SUPPORTING_FILES; expose CodegenConstants.GENERATE_SUPPORTING_FILES to templates.
• Update PythonClientCodegen templates (model_generic, model_oneof, model_anyof, api_client) to import the shared helpers when useLegacyModelHelpersModule is set; add _legacy_model_helpers.mustache.
• Fall back to inlined helpers when generating models only, keeping model-only outputs self-contained.
• Add focused tests for import vs inline behavior and regenerate the python-legacy-model-dictionaries sample; source size drops by ~7%.

^{Written for commit 0c1bee9. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 18 files

_{Re-trigger cubic}