[backport v2.8.next1] Steve socket is spammed on fresh visit to a resource detail page

[github-actions]

This is a backport issue for #10540, automatically created via GitHub Actions workflow initiated by @richard-cox

Original issue body:

We should tackle this after fix for rancher/rancher#41809 has merged

Setup

• Rancher version: 2.7.x

Describe the bug

After visiting by url / refreshing on an oldish steve based resource's detail page the steve socket gets spammed with messages.

This probably applies to the resource's edit page as well.

• Very roughly we fetch resources. That response contains a revision. We watch for changes for that set of resource over web socket using the revision as a starting point
• If that revision is too old we should get a resource.error too old message
• The UI then knows to make a fresh request for resources which will have the latest changes and the current reviewed
• There were scenarios where the too old message was not provided and will be / has been fixed via [SURE-7122] Excessive WebSocket activity when watching resources with permission by name rancher#41809
  • Socket spamming was due to a resource.stop --> resource.start --> resource.stop --> resource.start forever cycle
  • Fix means the cycle does not occurr resource.stop & resource.error 'too old' --> http request --> resource.start
• This revealed another issue related to spamming sockets to do with too old handling
• With the fix above fetched / watching resources is fine, however fetching / watching a single resource is not
  • Fetch single resource & use revision from within to watch (note - not the current system revision) --> resource.start with old revision --> resource stop & resource.error too old --> http request for single resource & use revision from within forever cycle
  • There's also a bug where the namespace for the single resource http request isn't passed through to the url construction, so the request is borked anyway

To Reproduce

• Bring up a system with the fix from [SURE-7122] Excessive WebSocket activity when watching resources with permission by name rancher#41809
• Navigate to a configmap page that's pretty old
• Bring up dev tools --> network tab
• Refresh the page
• View the socket messages for the ds cluster's socket connection

Result

• Multiple too old messages as the ui tries to watch with the old revision AND http requests with a bad url (no ns)

Expected Result

• socket message with valid revision (http requests TBD)

[izaac]

Reproduced on v2.8.1

Steps here

2024/03/15 18:08:41 [DEBUG] WatchNames received error: event watch error: too old resource version: 37058 (45171)
2024/03/15 18:08:41 [DEBUG] WatchNames received error: event watch error: too old resource version: 37058 (45171)
2024/03/15 18:08:41 [DEBUG] WatchNames received error: event watch error: too old resource version: 37058 (45171)
2024/03/15 18:08:41 [DEBUG] WatchNames received error: event watch error: too old resource version: 37058 (45171)
2024/03/15 18:08:41 [DEBUG] WatchNames received error: event watch error: too old resource version: 37058 (45176)
2024/03/15 18:08:41 [DEBUG] WatchNames received error: event watch error: too old resource version: 37058 (45178)
2024/03/15 18:08:42 [DEBUG] WatchNames received error: event watch error: too old resource version: 37058 (45178)

[izaac]

I wasn't able to reproduce it on v2.8.3-rc4

And the configmap request spam described at the bottom of this comment on the parent issue of this backport is not reproducible either.