Crash-loop container detection

v0.17.0-beta

Crash-loop container detection. The periodic health report now includes each container's Docker restart counter (State.RestartCount). A container crashing and restarting in a tight loop (on a missing mandatory environment variable, for instance) could previously appear continuously "Running" on the hub side, because each poll (every 30s) had a good chance of catching it between two crashes. The hub now compares that counter to the value observed on the previous round: if it went up, the container is marked failed even if it happens to be running at the moment of the report. Technical groundwork on the agent side only (the real detection fix is on the hub side); costs a single extra docker inspect call per tick instead of two (merging the former StartedAt/HEALTHCHECK pair).