Improve V3 Conversation Performance

[nplasterer]

Benchmark and add local mem caching to speed up performance.

[neekolas]

What exactly would we be caching? The V2 mem cache is for the conversations, which are already stored in the V3 DB

[nplasterer]

So I did a little digging to try and get to the bottom of where the performance issues truly are. This is the side by side comparison of iOS proper and iOS RN. It looks like syncing is just going to be 2x slower than loading from V2 was. Something we would need to improve on the libxmtp side. As for loading it looks like loading in V3 is 3x faster than loading from the mem cache in iOS proper but 3x slower in RN. So something is having trouble serializing across the bridge.

User	Action	iOS Time (ms)	RN Time (ms)
Alix	loaded 50 v2Convos	122.92	120
Alix	2nd loaded 50 v2Convos	9.14	13
Davon	synced 50 Dms	307.52	240
Davon	loaded 50 Dms	3.78	30

[nplasterer]

So I think the culprit is the peerInboxId. I think we need a join table of sorts in libxmtp to make this look up a lot less costly.

[nplasterer]

xmtp/libxmtp#1196