network: batch transactions on broadcast

[fyrchik]

Closes #665 .

[codecov]

Codecov Report

Merging #681 into master will decrease coverage by 0.19%.
The diff coverage is 0%.

@@            Coverage Diff            @@
##           master     #681     +/-   ##
=========================================
- Coverage   65.22%   65.03%   -0.2%     
=========================================
  Files         125      125             
  Lines       10743    10775     +32     
=========================================
  Hits         7007     7007             
- Misses       3457     3489     +32     
  Partials      279      279

Impacted Files	Coverage Δ
pkg/network/server.go	16.38% <0%> (-1.2%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b5ec94b...a271b86. Read the comment docs.

[codecov-io]

Codecov Report

Merging #681 into master will decrease coverage by 0.24%.
The diff coverage is 2.27%.

@@            Coverage Diff             @@
##           master     #681      +/-   ##
==========================================
- Coverage   65.96%   65.72%   -0.25%     
==========================================
  Files         128      128              
  Lines       10992    11033      +41     
==========================================
  Hits         7251     7251              
- Misses       3459     3500      +41     
  Partials      282      282

Impacted Files	Coverage Δ
pkg/network/server.go	15.15% <0%> (-1.34%)	⬇️
pkg/core/mempool/mem_pool.go	97.09% <100%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ff551d2...cbf26ba. Read the comment docs.

[fyrchik]

Some results on benchmarking:
I have tried neo-bench with different settings. Everytime RPS was higher after this patch.

Noticeable changes: after increasing mempool size to 500k, this bench showed interesting results.
.make/runner.sh -f $(DC_GO_IR) -f $(DC_GO_RPC) -i /dump.txs -d "Go4x1" -m rate -q 1000 -z 5m -t 30s -a go-node:20331

Before the patch (on the current master) amount of tx per block was roughly the same (8k -- 12k) with new blocks accepted ~every 20s.
After the patch amount of tx per block has been steadily increased during the bench up to 30k tx and 30s per block. Because we propose ALL currently verified transactions in a block, this can show that transactions become verified faster.

[roman-khimov]

From what I see in my testing of it, RPS is improved and block generation is changed a bit in that it really tends to produce bigger and bigger blocks over time. But at the same time I don't see any clear negative effect on TPS, so even if it just improves RPS it's a good change.