Avoid bad timing on posits

[jakmeier]

For presignature and triple generation, the proposer and deliberator roles time out at the same time.

The proposer tries to move forward on a timeout, if the threshold has been met. But a deliberator will simply abort. This leads to bad timing and non-determinism if even just a single node is not responding.

Increasing the deliberator timeout resolves the problem.

This also increases the frequency for checking the expiration, to speed up the process.

[jakmeier]

@ChaoticTempest This is what I mentioned in today's meeting. I just saw a test (test_sign_contention_5_nodes) failed, I will need to look into that.

[jakmeier]

Unfortunately tests are non-deterministic.

test_sign_contention_5_nodes and test_presignature_timeout sometimes fails in CI

locally, I got test_sign_contention_5_nodes failing once but mostly it runs without issues.

I probably won't have time to dig deeper today. But we shouldn't merge this while the issue remains.

[jakmeier]

I had to change the delay for deliberator expiration to make this work reliably.

In my initial fix, I used 2x the proposer timeout for deliberators. This makes it much more likely that round 1 of posits gets accepted. But I didn't consider that for round two, this creates another (perfectly) bad timing. 2x is pretty much the worst choice for the delay. Thus, resulting in the non-deterministic test failures I had.

Now I use a fixed 2s extra for deliberators to expire. That should be enough to receive the proposer START message in time. Combined with a proposer expiration of 10s, also leaves enough time for the next round to conclude.

[volovyks]

@ppca cases::mpc::test_sign_contention_5_nodes is still failing, often, so the issue is relevant
cases::mpc::test_sign_requests_wait_for_presignatures is unstable as well