Fix Stake Distribution done multiple times per epoch

[jpraynaud]

Issue

In case of error in the state machine or when the signer/aggregator node restarts, it computes the stake distribution which is a very memory/compute intensive operation even though it has already been recorded previously for the same epoch.

This leads to multiple problems:

• Long startup times
• Very painful bootstrap of networks as the open window for registration is very narrow

To do

Retrieve the stake distribution if and only if it is not already recorded for the epoch