feat(client): count sleap-app and sleap-nn processes in session metrics

[gitttt-1234]

Stacked on #509 (fix/tofu-backend-bucket-region) so the CLI fix and this change can be installed from one branch. Merge #509 first; GitHub will then retarget this PR to main.

Summary

Session metrics' time-to-first-label/train/track milestones only recognise SLEAP 1.x processes (sleap-label, sleap track …, python -m sleap.cli train …). On a sleap-app deployment they never fire. This maps sleap-app's processes onto the same canonical names:

Process (as seen in /proc/<pid>/cmdline)	Counted as
/usr/bin/sleap-app (GUI binary, from the .deb)	sleap-label
<uv python> ~/.local/bin/sleap-nn train …	sleap-train
sleap-nn predict … (or track on sleap-nn < 0.2)	sleap-track

The DB columns, cohort summary, export and process_allowlist default don't change. Removing a canonical name from the allowlist still turns off its sleap-app equivalent.

How sleap-app launches these

From talmolab/sleap-app main at 3010697:

• The GUI binary is the Cargo package name, sleap-app.
• Training: runTraining → run_python_command("sleap-nn", ["train", …]) (src/platform/backend.ts).
• Inference: sleap-nn predict on sleap-nn ≥ 0.2, otherwise sleap-nn track (pickInferenceSubcommand).
• sleap-nn export and the overlay/progress sidecars (<python> -u <script>) aren't milestones and don't match.

The other session metrics need no changes for sleap-app:

• The window title <file> - SLEAP v<version> matches the default sleap pattern.
• .slp files are HDF5 with a frames table.
• Models go in <project>/models/<run>/, so **/models/**/training_log.csv finds them.

Testing

• Added tests in packages/client/tests/monitoring/samplers/test_processes.py for the GUI binary, sleap-nn train launched via uv, predict/track, unknown subcommands, and allowlist handling.
• PYTHONPATH=src uv run pytest in packages/client: 186 passed. ruff check packages/client: clean.
• Not yet checked on a live sleap-app VM.

🤖 Generated with Claude Code