Feat: Add watcher support indexing multiple homeservers

[amirRamirfatahi]

Fixes #200 and #209

Pre-submission Checklist

For tests to work you need a working neo4j and redis instance with the example dataset in docker/db-graph

• Testing: Implement and pass new tests for the new features/fixes, cargo nextest run.
• Performance: Ensure new code has relevant performance benchmarks, cargo bench

[SHAcollision]

I thought about storing Homeserver into Neo4J also when I implemented it. I think the main issue with of storing Homeserver in the graph is a possible dataloss of Redis indexes on crash.

In case of restart/crash, we might restore the cursor from graph instead of Redis. Redis last snapshot might not be the latest cursor on graph, therefore, we will resume watching homeservers and skip some events.

I'm sure we can mitigate this easy by only restoring Homeserver from Redis on watcher startup, without fallback to graph (or only on some conditions).

[amirRamirfatahi]

Hmmm...I don't know if I understood that correctly or not, but I'm also updating the cursor in graph?

[SHAcollision]

Yes, and that's exactly the problem: the cursor in graph might be newer than the latest synced events in Redis in case of a restart/crash.

In other words, our assumption has always been that using Redis without persistence is OK and safe if and only if when we lose Redis data we also lose the cursor. This way, in the next startup, the lost events will be re-acquired. However, saving the cursor to graph breaks this guarantee.