soroban-rpc: Add in-memory event store backed by circular buffer

[tamirms]

What

Add in-memory event store backed by circular buffer. This code is taken from #355 . To minimize merge conflicts, the ingestion and DB code changes will be included in a separate PR.

Why

This is necessary to complete stellar/go#4718

Known limitations

[N/A]

[2opremio]

As I mentioned before, I would consider some modifications to the data structure so that we can (mostly) allocate it in advance

Also, it would be good to know how much memory could be consumed by this (e.g. assuming full ledgers and certain amount of events per operation)

[tamirms]

I started looking into @2opremio suggestion to take advantage of the fact that there are at most 1000 operations per ledger. I used the following representation:

type bucket struct {
	ledgerSeq uint32
	ops  []opEvents
}

type opEvents struct {
	events  []xdr.ContractEvent
	txIndex uint32
	opIndex uint32
}

// MemoryStore is an in-memory store of soroban events.
type MemoryStore struct {
	lock sync.RWMutex
	// buckets is a circular buffer where each cell represents
	// all events occurring within a specific ledger.
	buckets []bucket
	// length is equal to the number of ledgers contained within
	// the circular buffer.
	length uint32
	// start is the index of the head in the circular buffer.
	start uint32
}

but the code started to be more complex and less readable because of the fact that only operations per ledger are fixed (not transactions). I'm going to file an issue to investigate improving this representation in the future

[tsachiherman]

Few small comments, but looks great otherwise.

[tamirms]

@tsachiherman can you take another look? I just pushed a commit which addresses your latest round of feedback

[tsachiherman]

looks great! thanks for the changes.