fix(network): increase state concurrency and syncer lookahead

[teor2345]

Motivation

On my machine, the syncer can get 4000-6000 blocks ahead of the state. This causes the sync downloader to hit its lookahead limit, and return an error.

This bug happens a lot on testnet, and occasionally on mainnet.

These logs and stalls make it hard to test the other code I'm working on. (And they are a usability issue.)

Solution

• increase the state buffer bound to the maximum of the sync, inbound, and mempool buffers
• increase the sync downloader lookahead limit to 6000 blocks by default

This is a temporary quick fix to #3438, but it doesn't close that ticket.

With these changes, my testnet instance syncs in 1 hour, and my mainnet instance syncs in 6 hours. This fix significantly reduces the number of syncer stalls on both networks.

Review

@jvff can review this PR.

I am looking for a quick review to make sure these temporary changes don't break anything.
I expect we will make more changes when we actually schedule #3438 in a sprint.

Reviewer Checklist

• Code implements Specs and Designs
• Tests for Expected Behaviour
• Tests for Errors

Follow Up Work

[codecov]

Codecov Report

Merging #3455 (fac5caf) into main (499ae89) will decrease coverage by 0.02%.
The diff coverage is 75.88%.

@@            Coverage Diff             @@
##             main    #3455      +/-   ##
==========================================
- Coverage   78.34%   78.31%   -0.03%     
==========================================
  Files         267      272       +5     
  Lines       31526    31549      +23     
==========================================
+ Hits        24698    24707       +9     
- Misses       6828     6842      +14