Automate performance regression measurements for the long-chat fixture

[throwingogo-hub]

Turn the existing offline long-chat measurements into a repeatable maintainer command that reports DOM query counts and observer-callback timing without adding runtime dependencies to the extension.\n\nAcceptance criteria:\n- Use test/mock.html only; do not automate a real private conversation.\n- Produce stable, clearly labeled fixture measurements.\n- Document machine/browser variability instead of promising a universal speedup.\n- Keep the published extension package limited to the current runtime files.

[ahmdkaml]

Hi! I think I may have an approach for this issue, but I'd like to check whether it matches what you have in mind before I start implementing it.

My rough idea is to keep everything based on the existing test/mock.html fixture and add instrumentation that is only used for the benchmark. Rather than changing the extension's runtime behavior, the instrumentation would collect measurements such as DOM query counts and MutationObserver callback timing while the fixture runs.

To make it repeatable, I'd add a maintainer command that launches the fixture, runs the benchmark, and prints the collected measurements in a clear, labeled format. Any tooling needed to run the benchmark would remain development-only, so the published extension package would continue to contain only the current runtime files.

For completion, I was initially thinking of using a simple mechanism to detect when the benchmark has settled before printing the results, rather than building a more complex synchronization system. If you have a preferred way to determine when the fixture has finished, I'm happy to follow that instead.

Before I start working on this, does this general direction fit what you were expecting for the issue, or is there a different approach you'd prefer?