Fix issues with supervisor startup timeout

[jmcphers]

This change addresses two issues with the kernel supervisor's startup timeout measurement.

First, the timeout countdown starts when we ask for the terminal to be created, not when the terminal is actually established. When the supervisor extension starts early in the boot process, the terminal infrastructure/ptyhost might not be fully online, so the createTerminal can spend a long time waiting for it to be ready, exhausting most or all of the timeout.

Second, we retry every 100ms to find the connection file, but don't try more than 100 times no matter what that timeout is set to, with the result that you can only ever wait for 10s once we get into the main retry loop.

There are four fixes in this PR:

• Use two distinct timeouts, one for the terminal startup and one for the supervisor startup. Don't start the supervisor's countdown until the terminal is established and we have a PID. For simplicity, these timeouts are both controlled by the same setting.
• Don't limit retries; keep trying every 100ms until we have exhausted the startup timeout.
• Increase the startup timeout to 15s, as it seems that 10s isn't enough time to wait for the terminal to start in some configurations.
• Add more logging to help us understand how much time is spent in terminal startup and how much is spent in supervisor startup.

Note that all of these changes are speculative as I have not been able to reproduce the error myself; they are based on analysis of logs provided by @jennybc.

Addresses #11010

Release Notes

New Features

• N/A

Bug Fixes

• Address issues with startup timeouts when starting kernel supervisor (Console can not start after version 2025.10.1 #11010)

QA Notes

Jenny noted that this happens specifically when updating to a new version of the daily build.

[github-actions]

E2E Tests 🚀
This PR will run tests tagged with: @:critical

^{readme  valid tags}