TestF3LateBootstrap is flaky

[masih]

Over the course of latest development in this repo TestF3LateBootstrap is pupped up a number of times as being flaky. It's timeout has already been bumped to 1m, and it is the slowest test we have in this repo.

One cause that I have observed is that phases time out before messages are delivered which results in repeated base decision and causes assertions that check if a certain epoch is finalised take too long. A root cause is the way mock clock is used in the tests: it is used to grow the chain and gpbft. Another issue is the way clock advancing works relative to assertion checks. As is the test is very CPU hungry as it runs assertions in a hot-loop.

I captured this as a separate issue from #818 to emphasise the need for making this test run faster as it slows down development.