(bug) fixes various bugs and performance issues

[gianlucam76]

Bugs

• Corrected a logging mistake where the cluster name was accidentally populated with the cluster namespace in two places, making log lines useless when debugging by cluster name.
• Fixed wrong control flow in the pull-mode agent status handler: one error type was being checked in the wrong branch (it was only reachable when there was no error, so it never fired), and the function could panic with a nil dereference when the agent returned an error without a status payload.
• Fixed a data race in the Helm chart manager: one method was reading from a shared map without holding the mutex, while concurrent reconcilers could be writing to it.

Performance

• The dependency manager's background update loop was holding a write lock — which blocks all concurrent reconcilers — across a batch of API calls. It now snapshots what needs updating, releases the lock, makes the API calls, then re-acquires the lock only to clear the completed entries.
• The dependency manager's startup rebuild loop was spinning at full speed retrying on any API error with no delay, which could hammer the API server during an outage. It now waits 5 seconds between retries and respects context cancellation.