Handle filesystem surrogate characters in test node ID hashes

[xiehuanyi]

A Linux test filename containing an undecodable filesystem byte produces a surrogate-containing node ID. With pytest's cache provider disabled, ordinary pytest runs it successfully, but pytest-randomly aborts collection while hashing the node ID with strict UTF-8 encoding.

Use UTF-8 surrogatepass for the hash input, preserving the encoding and order for existing ordinary Unicode IDs. Add an end-to-end filename regression and a changelog entry.

Validation: the new regression fails before and passes afterward. All 40 tests pass, including with the warning-as-error settings from tox. All applicable repository-pinned pre-commit checks pass.

Scope: pytest's separate cache serialization limitation still requires -p no:cacheprovider for such filenames. The filesystem regression runs on Linux.

Prepared and verified with Codex (GPT-6.1-sol) assistance.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.