Optimize run intervals handling in signer and aggregator state machines

[jpraynaud]

Why

At the end of a cycle in the state machines of the signer and aggregator, there is fixed delay before the next cycle. We need to modify the computation of the delay to start a new cycle as soon as the run interval has elapsed since the beginning of the last cycle.

What

Modify the run interval implementation in the state machines of the signer and aggregator.

How

• Use the interval function from tokio
• Make sure there is no problem to run a new cycle immediately after another cycle (without delay)
• Run a successful aggregator stress test in the CI
• Make sure the Hydra CI is not flaky