Collector onboarding/deployment don't show input-related issues after initialization

## Expected Behavior
If the configured collector ingest input is in a failed state (e.g. due to port collision), a user eligible to enroll collectors should have some kind of warning about this on the deployment pages (either onboarding or deployment).

## Current Behavior
During initial onboarding there is a flow to set up the ingest input for the first time (sets the collector config and starts the input). However, if the input fails, the user navigates away and comes back to the onboarding flow, the fact that the input is not actually usable is no longer clearly visible to the user.
It is reflected in the Cluster Health overview, but the onboarding itself pretends everything is fine (because it only looks at the persisted collector configuration).

## Possible Solution
Also check the input state and if it's not `RUNNING` then we should show the `IngestEndpointStrip` like during initial onboarding.

## Steps to Reproduce (for bugs)
1. Go through collector onboarding with an admin account.
2. Set up the ingest input using a port you know will conflict, e.g. the REST API port.
3. The endpoint configuration strip will show that it failed.
4. Navigate away and come back to the onboarding
5. The warning no longer shows up, this is the bug.

## Your Environment

* Graylog Version: 7.2.0-beta.2
* Java Version:
* OpenSearch Version:
* MongoDB Version:
* Operating System:
* Browser version:

## Checklist
[] This issue fix need to be backported.
[] Does this issue have **security** implications?

Start with the collector onboarding and deployment pages, especially the IngestEndpointStrip, and compare their handling with the Cluster Health overview. Reproduce the failed input using a conflicting port, then trace whether the current input state is checked after returning to onboarding. Done means a non-RUNNING input shows the warning again.

Written by the indexing model from the issue text.