fix(tui_gateway): the change watcher also probes served sibling profiles for cron/platform edits

[EloquentBrush0x]

Summary

change_watcher.py::_sessions_sig() already probes every served sibling profile home (_served_profile_homes) alongside the launch home, per its own comment: "a backend serving several profiles owns one store per profile, so every served sibling home is [probed]" — otherwise a routed profile's Bot Chat never refreshes (#99333).

cron.changed and platforms.changed never got the same treatment, even though the files they watch are just as much per-profile-home artifacts as state.db:

• cron/jobs.py's own docstring: a named profile's jobs live at ~/.hermes/profiles/<name>/cron/jobs.json.
• hermes_cli/container_boot.py:212 writes gateway_state.json per profile_dir.

In a desktop backend hosting 2+ named profiles, a secondary profile's cron edit or platform connect/disconnect never broadcast a .changed event — only the launch profile's own file moves were ever seen.

Changes

• Added _home_and_siblings_mtime_ns(*parts), mirroring _sessions_sig()'s pattern of joining the same relative path under _watcher_home() and every entry of _served_profile_homes.
• cron.changed and platforms.changed now use it instead of the launch-home-only _home_mtime_ns.

Scope note — two watches deliberately left untouched

• pet.changed: its payload function (_pet_changed_payload) reads only the launch profile's own active pet. Extending just the signature function to notice a served profile's pet change would broadcast an event to every client, but the payload would still describe the launch profile's pet — a data-correctness problem, not something a signature-only fix should paper over. Left out of scope.
• bot_relay.outbox.pending: verified this does not have the same gap before touching it. get_profile_dir() always resolves named profiles under <default home>/profiles/<name> (hermes_cli/profiles.py), and _bot_relay_outbox_sig()'s own root-collapsing logic (home.parent.parent if home.parent.name == "profiles" else home) already folds every such profile home back to the shared launch root that _watcher_home() covers. Adding _served_profile_homes there would only re-probe the same directory. Confirmed by tracing _get_profiles_root() rather than assuming the pattern generalizes.

Test plan

• Added test_served_profile_cron_jobs_move_broadcasts_cron_changed and test_served_profile_gateway_state_move_broadcasts_platforms_changed, mirroring the existing test_served_profile_store_move_broadcasts_sessions_changed harness (real server._profile_home() call, not a hand-inserted set entry).
• Mutation-verify: reverted the production file only, confirmed both new tests fail (assert (...) in []); restored the fix, all 18 tests in test_change_watcher.py pass.
• ruff check clean on both touched files.

🤖 Generated with Claude Code