runtime: make Slack bridge restart policy adaptive (backoff/jitter/failure threshold)

[benvinegar]

Problem

start.sh runs the Slack bridge in a while true loop with a fixed 5s restart delay. Persistent faults can produce noisy restart storms and ambiguous operational state.

Proposed solution

• Replace fixed-delay infinite restart with adaptive policy:
  • exponential backoff (+ jitter)
  • reset backoff after stable run window
  • max-consecutive-failure threshold with clear alert/status signal
• Emit structured restart reason/attempt logs for easier triage.
• Keep current behavior as fallback if policy env vars are unset.

Helpful context

• start.sh launches tmux session slack-bridge and restarts bridge with hardcoded 5s sleep.
• Broker health file exists (broker-health.json) but restart policy is still simplistic.