bug: Race condition in Docker manager GetDeployment

[nfebe]

Problem

internal/docker/manager.go (lines 58-72) releases the read lock after GetDeployment but before populateContainerInfo. Container state can change between lock release and the subsequent read, returning stale or inconsistent data.

Impact

Intermittent stale container status, especially under concurrent API requests or rapid container state changes (start/stop/restart).

Proposed Fix

Hold the lock through the entire read-populate cycle, or use a snapshot/copy pattern so the returned data is consistent even if the underlying state changes.