delay clearing instance cache when an instance state change fails

[edlerd]

Done

• delay clearing instance cache when an instance state change fails
• avoid reporting wrong intermediate instance status as returned by the api shortly after a state change failure
• related to https://github.com/canonical/lxd/issues/14042

QA

1. Run the LXD-UI:
   • On the demo server via the link posted by @webteam-app below. This is only available for PRs created by collaborators of the repo. Ask @mas-who or @edlerd for access.
   • With a local copy of this branch, build and run as described in the docs.
2. Perform the following QA steps:
   • create an instance, set "Resource Limits" > "Max number of processes" to 1.
   • Start the instnance
   • Ensure the instance is reported as stopped, not as running on the startup error.

[webteam-app]

Demo

Jenkins

demos.haus