perf/tests: use more appropriate data type for package filter policy packages and add tests

[dhimasardinata]

PackageFilterPolicy normalizes installer.no-binary, installer.only-binary, and related package-filter settings once, but allows() and has_exact_package() previously scanned the normalized sequence each time they were called.

Chooser policy checks run for every candidate link, so this stores the normalized package names as an immutable frozenset for repeated membership checks.

A small local benchmark over a 1,001-entry normalized package policy:

old=0.234734s
new=0.006853s
speedup=34.3x

Pull Request Check List

• Added tests for changed code.
• Updated documentation for changed code. N/A, no user-facing behavior change.

Local checks:

PYTHONPATH=src uv run --with pytest --with pytest-xdist --with responses --with pytest-mock --with deepdiff python -m pytest tests/config/test_config.py tests/installation/test_chooser.py -q
uv run --with ruff python -m ruff check src/poetry/config/config.py tests/config/test_config.py
uv run --with ruff python -m ruff format --check src/poetry/config/config.py tests/config/test_config.py
PYTHONPATH=src uv run --with poetry-core --with packaging --with mypy python -m mypy src/poetry/config/config.py
git diff --check

[sourcery-ai]

Hey - I've left some high level feedback:

• Since packages and _package_set must stay in sync, consider making packages immutable (e.g., a tuple) or otherwise preventing external mutation, so bugs aren’t introduced by code that modifies packages without updating _package_set.
• If the policy is never mutated after initialization, using a frozenset instead of set for _package_set could better express immutability intent and slightly reduce the risk of accidental modification.

Prompt for AI Agents

Please address the comments from this code review:

## Overall Comments
- Since `packages` and `_package_set` must stay in sync, consider making `packages` immutable (e.g., a tuple) or otherwise preventing external mutation, so bugs aren’t introduced by code that modifies `packages` without updating `_package_set`.
- If the policy is never mutated after initialization, using a `frozenset` instead of `set` for `_package_set` could better express immutability intent and slightly reduce the risk of accidental modification.

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}

[dhimasardinata]

Addressed Sourcery feedback in a0a8f85 by storing normalized packages as an immutable tuple and using a frozenset for repeated membership checks.

[radoering]

A small local benchmark over a 1,001-entry normalized package policy:

I doubt that real numbers become that high. I assume that 100 is quite much and most of the times it will be less than 10. Nevertheless, frozenset feels like the more appropriate data type.