[unsupervised AI] Isolate CPU-time GC logging from RSS diagnostics

[YusefSyed]

Warning

This PR was written autonomously by an AI agent and has not been reviewed
by a human yet. Maintainers should ignore it until the human author has reviewed,
understood, and approved everything that the AI agent wrote.

test_gc_diagnosis_cpu_time asserts that no INFO message is emitted while full GCs consume less than 75% of CPU time. However, it leaves the independent RSS threshold at its 10 MB default, so a sufficiently large memory release emits an INFO message even when the CPU assertion is satisfied. A Windows Python 3.11 run failed this assertion after a GC released 12.86 MiB, exceeding the 9.54 MiB RSS threshold.

Set the RSS threshold to infinity for this CPU-specific test. Finite RSS releases then remain at DEBUG, while CPU INFO logging keeps its existing 75% threshold. Production defaults and the separate RSS test are unchanged.

Validation on macOS ARM64 with Python 3.12:

• pytest -q distributed/tests/test_gc.py: 2 passed, 1 xfailed (the existing flaky RSS test).
• CPU test in five separate pytest processes: all passed.
• A controlled callback probe with a 20 MB RSS release reproduced the unwanted RSS INFO at 50% CPU under the default threshold, produced no INFO with the infinite RSS threshold, and still emitted CPU INFO at 90% CPU.
• Pre-commit hooks for distributed/tests/test_gc.py: Ruff check/format and Mypy passed; git diff --check passed.

[github-actions]

Unit Test Results

See test report for an extended history of previous test failures. This is useful for diagnosing flaky tests.

    40 files  ±0      40 suites  ±0   14h 44m 39s ⏱️ + 25m 41s
 4 160 tests ±0   3 981 ✅ +1    178 💤  - 1  1 ❌ ±0 
80 961 runs   - 1  76 720 ✅ +1  4 240 💤  - 2  1 ❌ ±0 

For more details on these failures, see this check.

Results for commit ef9668b. ± Comparison against base commit dc182bd.