[ciltocruz]

Fixes #188

Problem

When a container is recreated (Watchtower, docker-compose up -d, podman restart, ...), each monitor keeps the Docker API handle for the old container id. The new container has the same name but a new id, so:

• The poll loop logs Container not available anymore (3a) (DockerError 404 ...) every retry, forever, and never recovers (worst case through docker-proxy where events can be missed).
• If the destroy event is missed, the later create event hits Container already monitored and is dropped, so the stale monitor is never replaced.

Also seen with podman: a stopped then restarted container stays unknown until HA is restarted.

Fix (custom_components/monitor_docker/helpers.py)

1. On DockerError 404 in the container poll loop, try to re-attach by name (_handle_container_gone). If the recreated container is found, monitoring continues transparently with the fresh handle (Re-attached to recreated container).
2. If the container name no longer exists (2 consecutive 404 re-attach failures), the container monitor asks the DockerAPI instance to drop its own monitor (on_container_gone), so entities are removed and a later create event re-adds everything cleanly.
3. _container_add now detects the name is monitored, checks the new container's id, and replaces the stale monitor instead of bailing out with Container already monitored.

Behavior notes

• Generic (transport/timeout) errors keep the old 40s retry behavior unchanged.
• No config or UI changes; sensors/switches/buttons are recreated normally by the existing load_platform flow when a monitor is re-added.