fix(worker): prevent long internal parser filenames

[suguanYang]

Summary\n\n- cap internal parser basenames at 240 bytes before moving downloaded files\n- preserve the extension and append a deterministic hash when truncating\n- add a regression test for URL-encoded long filenames\n\n## Incident\n\nProduction trace 01a056b7ab03e18328f0d935beff61c3 failed job job_9acad6bd015f with OSError: [Errno 36] File name too long while renaming a temporary PDF. The source filename was URL-encoded Chinese text and exceeded Linux's per-component filename limit.\n\n## Verification\n\n- uv run ruff check on the touched files\n- uv run --project apps/worker pytest apps/worker/tests/unit/test_internal_parse_name.py -q\n- uv run pyright on the touched files\n\nAll focused checks pass.

[suguanYang]

Addressed both CodeQL comments in commit 93d9eee: file existence/content are computed before assertions, and the test now uses Path.read_bytes() instead of an unclosed open(). Focused Ruff and pytest checks pass; GitHub checks are rerunning.