GRPC ingester config, code gen, wiring and docs

[p0mvn]

Partial merge of: #82

This PR focuses on merging the foundational parts with low risk. The new ingester component and server are disabled by default.

GRPC Ingest Protocl Documentation

This is a component that is responsible for ingesting and processing data.

It exposes a GRPC API for Osmosis node clients to send data to be ingested.

sequenceDiagram
    participant NodeIngester
    Note over SQS,NodeIngester: SQS Ingest RPC Communication
    Note over NodeIngester: Acquires lock on block processing
    NodeIngester->>SQS: StartBlockProcess(height, taker_fees)
    Note over SQS: SQS starts processing height
    SQS-->>NodeIngester: StartBlockProcessReply
    NodeIngester->>SQS: ProcessChainPools(stream)
    NodeIngester-->>SQS: ProcessChainPools.Send(pools_data_chunk)
    Note over SQS: SQS transforms and loads pools into ingester cache
    NodeIngester-->>SQS: ProcessChainPools.Send(pools_data_chunk)
    Note over SQS: SQS transforms and loads pools into ingester cache
    NodeIngester->>SQS: ProcessChainPools.CloseSend()
    
    NodeIngester->>SQS: EndBlockProcess(EndBlockProcessRequest)
    Note over SQS: SQS commits all state into SQS repositories, making them available for clients.
    SQS-->>NodeIngester: EndBlockProcessReply
    Note over NodeIngester: Releases lock on block processing

Loading

Note that, as of right now the protocol is syncronous where each GRPC call happens in sequence. However, from the
node perspective it is processed in a separate goroutine, letting the node continue producing blocks. The node
acquires a lock on the block processing so that the interaction is not affected by synching.

This is not a concern since, when the node is caught up, the block time is approximately 4.5 seconds while entire
protocol is capped at 1.5 seconds.

Currently, we push all pool data into SQS every processed block. As we lower block time, we will introduce a mechanism for
pushing the pool data only for the modified pools. This will allow us to drastically lower the protocol interaction from 1.5 seconds.

Alternative methods will include making the protocol more asyncronous but this will require introducing more complex
locking mechanisms which are overkill today.

[p0mvn]

Note: 16 KB. We can't leave comments. Once we implement yml config, will be able to transcribe this better

[p0mvn]

Merging per discussions with Nicco. Will have a walkthrough on full completion