fix(http-client-python): preserve customized pyproject.toml fields

[l0lawrence]

Fixes microsoft#10311

Problem

The Python emitter regenerates pyproject.toml on every emit and clobbers manual edits to fields it doesn't own — for example downcasing "AI" to "Ai" in the description, dropping Python versions from the classifiers list, and replacing a custom repository URL with the generic Azure SDK one.

Fix

Extend the existing KEEP_FIELDS preservation mechanism so the emitter preserves the following from an existing pyproject.toml instead of overwriting them:

• project.description
• project.classifiers
• [project.urls]

On first emit (no existing file) these are still generated as before. When the file already exists, the user's values are carried through.

Changes

• general_serializer.py: extract description, classifiers, and urls into KEEP_FIELDS.
• pyproject.toml.jinja2: use preserved values when present, fall back to generated defaults otherwise.
• Added unit tests in tests/unit/test_pyproject_keep_fields.py.
• Added a changeset entry.

Validation

• New unit tests pass.
• Template renders valid TOML in both the default and preserved-fields branches (verified by round-tripping through tomli).
• Black formatting passes with the repo config.

[github-actions]

All changed packages have been documented.

• ✅ @typespec/http-client-python

Show changes

@typespec/http-client-python - fix ✏️

Preserve manually customized description, classifiers, and [project.urls] fields in an existing pyproject.toml instead of overwriting them on regeneration.