Improving Mimir.Worker Efficiency

[Atralupus]

Currently, Mimir.Worker processes data by comparing the tip received from GQL with data stored in MongoDB.
To achieve this, it uses an ActionHandler to examine executed actions from the transaction list and a Handler to build state by checking diffs. However, the diff operation imposes significant CPU and network overhead on the headless server.

I propose switching to a method where we run Libplanet directly to fetch the state and compare it with diffs and transactions.
This approach, however, is resource-intensive, almost equivalent to spinning up a new remote headless instance. Since we only need to fetch the state, I believe using paev and state snapshots would be sufficient.

This is still in the idea stage, but I plan to conduct a POC and share the code once it’s ready.