Optimize fetching transcoder pool

Livepeer fetches all active orchestrators from the chain. It happens in the function TranscoderPool(), which is called in a number of scenarios:

• during the node startup
• before initializing the round
• during calling reward
• while using webserver by Livepeer CLI (here and in other places)

Fetching orchestrators happens sequentially. Each transcoder is fetched by calling the on-chain contract function GetNextTranscoderInPool(). This process already takes almost 2 min and it will grow linearly together with the number of active orchestrators.

We need to research and implement an optimization to this process. Some first ideas:

• Use Subgraph with indexed chain data
• Create a smart contract which stores the orchestrator pool on-chain
• Cache active orchestrator pool in the local DB

I suggest to first research the domain of solutions, describe and review them, and when we agree on the solution, finally do the actual implementation.