fix(api-server): prevent race condition in server restart process

[o0charlie0o]

Description

Fixes a race condition in the API server restart process where the newly started server could be killed by a lingering timeout handler from the previous shutdown sequence.

Changes

• Refactored killApiServer to use a single promise with proper cleanup
• Added cleanup function to remove exit listeners and clear timeouts
• Ensured exit listeners are set up before sending kill signal to prevent missing exit events
• Added proper logging for server shutdown states

Background

Previously, using Promise.race allowed both promises to continue executing even after one resolved, which could lead to the force-kill timeout executing after a new server had started. This fix ensures that only one shutdown path (graceful or force) can execute, and all listeners and timeouts are properly cleaned up.

Testing

To verify the fix:

1. Start the dev server
2. Make a change to trigger a server restart
3. Confirm that the server successfully restarts and remains running
4. Verify that multiple restarts work consistently

Fixes #11719

[Tobbe]

Thanks for the PR. I'll add it to my list of things to review