[CON-201] Remove QueueInterfacer + increase RS update concurrency

[theoilie]

Description

• Make updateReplicaSet job processor init audius SDK instead of using QueueInterfacer so it has no more stateful dependencies
  • Same for the fetch cNode endpoint to spId mapping queue and then remove remove QueueInterfacer entirely
• Add maxUpdateReplicaSetJobConcurrency env var to give the updateReplicaSet queue a concurrency of 15

Tests

Brought up local stack and made sure the fetch-c-node-endpoint-to-sp-id-map queue was fetching the right data using the newly initiated libs, and made sure replica set update jobs go through.

Monitoring - How will this change be monitored? Are there sufficient logs / alerts?

Monitor replica set update jobs in the bull dashboard and see if any new errors come up. They should have 15 active jobs at a time (instead of just 1 like before).

[dmanjunath]

im supporting of trying this change, the only thing i'm a bit concerned about is the extra load every job initializing libs will place on this. for example this will have to do a full discovery selection and make health check reqs to all nodes etc. could we pair this with a watered down version of libs to prune stuff we don't need?

[theoilie]

looks like the things we need are:

• contracts.UserReplicaSetManagerClient.updateReplicaSet
• ServiceProvider.autoSelectCreatorNodes
• ethContracts.getServiceProviderList

I traced through the various top-level config options, and discoveryProviderConfig was the only one we can get away with not passing for the above functions. that will avoid the heavy discovery selection at least