Fix dashboard post-process restart

[aryaethn]

The credit for this bug and its fix goes entirely to @moodmosaic. I just coded his patch proposal from #110 to speed things up. Only small additions on top: two lines wrapped to 79 chars (the replacement notice now says "post-process container"), extra behavioural tests, and a stale USAGE.md line fixed.

Summary

• Dashboard p now refuses to start post-processing only when the post-process container is actually running, so an exited container no longer blocks a restart (fixes Dashboard post-process restart issue #110).
• When an exited container is present, p says it will be replaced; launch.sh post-process already runs docker rm -f on it before starting.
• p passes the dashboard's SWARM_CONFIG to launch.sh post-process, so dashboards opened with a non-default config launch the right post-process.
• P (logs) still works on an exited container, so a finished run's logs stay reachable.

Changes

• dashboard.sh: add post_process_container_running; use it for the p guard and the [p] footer hint; print a replacement notice for exited containers; prefix the launch with SWARM_CONFIG="$CONFIG_FILE".
• tests/test_dashboard.sh: the static checks from Dashboard post-process restart issue #110, plus behavioural tests that load the real post_process_container_exists/_running from dashboard.sh with a stubbed container_state across running, exited, created, paused, not found, and none.
• USAGE.md: note that p replaces an exited run; fix the stale "Trigger via [P]" text, which has meant logs since the key swap in 85f60cd.
• CHANGELOG.md: entry under Unreleased.

Test plan

• ./tests/test.sh --unit: 13 files, 1381 tests pass (macOS).
• /bin/bash tests/test_dashboard.sh passes under bash 3.2 (163 tests).
• Against the unpatched dashboard.sh, 6 of the new checks fail, so they catch the bug.
• shellcheck -s bash --severity=warning dashboard.sh tests/test_dashboard.sh is clean.
• ./tests/test.sh --all.
• Manual: run post-process from the dashboard, let it exit, press s, then p; it should warn that the previous container will be replaced and start a new run.
• Manual: open the dashboard with SWARM_CONFIG=<non-default>.json, press p; the post-process uses that config.

The last three are unchecked because Docker wasn't available locally.

Co-Authored-By: Claude Opus 5.5 noreply@anthropic.com

[moodmosaic]

LGTM!

[moodmosaic]

This resolves #110.

[moodmosaic]

Thank you, @aryaethn!