Added an "aggressive paging" mode to PagedNoteDataSource

[mplorentz]

Issues covered

#1192

Description

The problem here was that in paginated lists we could get a full page of notes from relays, none of which would actually get displayed in the UI. For example a user might post 20 replies in a row without posting a root note. Because PagedNoteDataSource was using cellForRowAt() to tell the pager to load more data if cellForRowAt() stopped getting called then we stopped loading more data.

My fix was to detect when we are getting to the end of the list and go into "aggressive paging" mode where we start calling pager.loadMore() on a timer. Calling pager.loadMore() constantly isn't actually as bad as it sounds. Unlike RESTful paging where you request page 0, then 1, then 2, in Nostr we use a sliding time window. We are asking relays "give me the 10 most recent notes published before X date". And X date is the oldest event returned so far, and it's unique to every relay we have a connection with. So calling loadMore() just slides the time window further back and doesn't queue up more than a page's worth of events to download at any given time.

How to test

This is tricky because it depends on live data that could change. i suppose I could've crafted some but I actually just used Onigirl (npub18jvyjwpmm65g8v9azmlvu8knd5m7xlxau08y8vt75n53jtkpz2ys6mqqu3) from our Discover tab and as of right now it looks like it's still reproducible on their profile.

1. Open Onigirl's profile npub18jvyjwpmm65g8v9azmlvu8knd5m7xlxau08y8vt75n53jtkpz2ys6mqqu3
2. Scroll back and verify that data pages

[joshuatbrown]

This all makes sense and the code looks good. However, I'm seeing a lot of these console messages when I go to the profile and scroll to the bottom:

Notice from nostr.fmt.wiz.biz: [NOTICE, ERROR: too many concurrent REQs]
Notice from offchain.pub: [NOTICE, ERROR: too many concurrent REQs]
Notice from soloco.nl: [NOTICE, ERROR: too many concurrent REQs]
Notice from e.nos.lol: [NOTICE, ERROR: too many concurrent REQs]
Notice from nos.lol: [NOTICE, ERROR: too many concurrent REQs]

To be fair, I'm seeing some from main, too, but not nearly as many as I see in this branch. And this branch actually loads the old posts, whereas main does not. So it works, but maybe we can back off some of the requests a bit, or something?

[mplorentz]

@joshuatbrown great catch! I did not see those but they were really bad. I found that I was calling the wrong decrementSubscriptionCount() function 😭 the one on RelayService closes the socket properly and the RelaySubscriptionManger doesn't. I started trying to refactor it but it was going to be a lot of work. Right now the RelayService does all the actual writing to websockets and the RelaySubscriptionManager is just a bookkeeper. If I refactored RelaySubscriptionManager to handle the raw websocket communications then it would either have to duplicate code in RelayService or it would get really big and complex. Our relay comms are already too big and complex so for now I just wrote a comment about it and left the architectural problems for a bigger refactor project.

[joshuatbrown]

Looking for a couple more changes. This is much better -- no more issues with too many concurrent REQs.

[joshuatbrown]

Oops!

[joshuatbrown]

Suggested change

	await relayService.decrementSubscriptionCount(for: subscriptionID)
	relayService.decrementSubscriptionCount(for: subscriptionID)

[joshuatbrown]

Perfect!