Some epochs might not get truevals; fix this

[alexcos20]

Right now, we use a simple logic in main.py:

• get last block -> get epoch for a PredictionContract, then submit trueval for past epoch

But this poses a couple of problems:

• what if RPC connection drops for 10 mins?
• what if our sources (binance, kraken, etc) are failing for 10 mins?
• what if our code fails and docker is restarted?

This can lead to a lot of epochs without truevals, hurting the rewards.

Proposal:
We should use subgraph and always get a list of epochs, per each PredictionContract, without a submitted trueval. (We have epoch status). Then submit truevals for those epochs.