perf: batched event reception

[link2xt]

Alternative to #7795

[link2xt]

I have tested it on desktop, it works, events are frequently batched into 3 or 5 events for a single call. But no idea on actual impact on the performance, did not measure anything.

[WofWca]

I could not perform the tests under the same conditions so far, because we have some hacky event logging there, which doesn't work with this new JSON-RPC method.

https://github.com/deltachat/deltachat-desktop/blob/416f8f9e9e734f49de348c8a35db3c072155a741/packages/target-electron/src/deltachat/controller.ts#L114-L184

However, the delay is down to ~150ms without any logging (compared to ~1300ms in my MR).

I will try to fix the logging and then re-do the measurements.

[WofWca]

OK, I have fixed backend logging on DC Desktop and performed another 150 measurements on this patch. The average is 238 ms, with standard deviation of 89.97 ms.
Admittedly, there were ~3 <50ms outliers in my measurements, and I discarded them.

For good measure, I also re-measured the delay with the old non-batch API, but without backend logging, and got 558.81 ms on average and 196.16 ms of standard deviation over 37 samples.
With backend logging and the old non-batch API we are back to 1172.61 ms on average with std deviation of 155.76 ms over 24 samples.

I also tested this MR without backend logging, and the delay is at blazing 130 ms on average with std deviation of 6.10 ms!

So, no matter if logging is on or off, this is an improvement!

[WofWca]

So this is technically not a breaking change, but, since DC Desktop is using some hacks, it would need to get adjusted. Anyways, I already have the code ready for that.

I didn't look to closely at the Rust part of the code, but it looks sane, and seems to work. The JS code looks trivial.

[WofWca]

So this is technically not a breaking change, but, since DC Desktop is using some hacks, it would need to get adjusted.

Regarding this: I think I could make an MR that adds support for this batch API, before this Core MR is merged, so as to not get other devs who always use latest core with latest desktop angry.

[link2xt]

I added doc comment and limited the number of returned events to 100 per batch. It is unlikely ever at 100 events, just to have some limit on memory usage.