Parallelize handling events in Orchestrator Watcher

[leszko]

After each new round is started, for each active orchestrator, we cache its stake in the local DB. This operation is performed sequentially for each orchestrator. In the loop we call cacheOrchestratorStake().

Currently, for 90 active orchestrator, it takes around 3 minutes to execute, but the time will grow linearly together with the number of active orchestrators. Therefore we need to parallelize the cacheOrchestratorStake() execution.

It may be as simple as adding the go keyword and removing the roundMu lock, but it's better to double-check if all operations are thread-safe.