Propagate worker usage errors to controller

[lin-hongkuan]

Fixes #109.

When collection on a worker raises pytest.UsageError (for example because a requested path does not exist), the worker exits with status 4 but the controller currently only sees an empty collection and finishes as "no tests ran" with exit code 5.

This stores the original UsageError arguments on the worker output during collection, then has the controller re-raise pytest.UsageError when a worker finishes with ExitCode.USAGE_ERROR. That lets pytest print the original error message and preserve the usage-error exit code.

Validation:

• python -m pytest -p no:twisted testing\acceptance_test.py::TestDistribution::test_missing_path_usage_error -q
• python -m pytest -p no:twisted testing\acceptance_test.py::TestDistribution -q
• python -m pytest -p no:twisted testing\test_remote.py -q
• python -m ruff check src\xdist\remote.py src\xdist\dsession.py testing\acceptance_test.py
• python -m ruff format --check src\xdist\remote.py src\xdist\dsession.py testing\acceptance_test.py
• python -m mypy --ignore-missing-imports src\xdist\remote.py src\xdist\dsession.py testing\acceptance_test.py

I also manually verified python -m pytest -n2 MISSING now returns 4 and prints ERROR: file or directory not found: MISSING. A full local python -m pytest -p no:twisted -q run was attempted but exceeded 5 minutes before completing in this environment.

[RonnyPfannschmidt]

Use modern wrappers