chore: ~remove~ ignore flaky test re capacity, serialize prove generating tests

[drahnr]

https://midengroup.slack.com/archives/C08UTTDUMAS/p1772228752145459

The test is failing in CI in various of my PRs, yet I cannot reproduce it locally, nor is the test critical.

[Mirko-von-Leipzig]

The culprit is simply CI performance :/

I suspect what's occurring is that the proving is simply too slow on the CI runner, and that the proof jobs are therefore timing out after 1 minute.

This is the sort of thing where we could "cheat" by using dependency injection to replace the proving with something cheap. However I dislike mocking like that.

Before we merge this, we should see if there are alternatives to improve the CI performance somehow. We already have dev profile optimisations specifically for this -- I wonder if we can go further? Or maybe we do need to mock this.. I kinda like having this test :) Though it isn't critical. The other tests in this file also take 60s+ so it is an issue to address maybe..

[drahnr]

The current CI failure is fixed in #1727

[drahnr]

I ignored the test that does 3 concurrent proof generation events and added #[serial_test::serial] annotations to the other tests that also request proving