chore: Apply block thread priority

[sergerad]

Summary

Closes #2375.

benchmark-mixed -n 50000 -p 50 --storage-map-entries 64 --account-update-blocks 20 -r 16

Metric	lock-free	apply-block-priority	Δ (branch vs lock-free)
Reads/sec	83,770	45,723	−45%
Total reads	33.7M	11.6M	−66%
P50 read latency	90µs	229µs	2.5× worse
P95	687µs	927µs	1.3× worse
P99	1.26ms	1.53ms	1.2× worse
P99.9	4.1ms	5.1ms	1.2× worse
Avg block insertion	426ms	382ms	10% faster
Total seeding time	405s	256s	37% faster
Avg get-batch-inputs	2.0ms	1.8ms	~equal

This branch is much faster on the write side (256s total seeding vs 304s on next, 382ms avg block insertion) but the read side degraded substantially — roughly half the read throughput of next (45.7k vs 73.8k reads/sec) and ~2.7× worse P50 (229µs vs 83µs). That's consistent with what the branch does: giving the apply-block thread elevated priority starves the concurrent readers under load. Tail latency also regressed vs both columns (P99.9 of 5.1ms vs 4.1ms lock-free / 9.8ms next — still ahead of next there).

Changelog

changelog = "none"
reason    = "Internal change only."

[Mirko-von-Leipzig]

Could you rerun the benchmarks you did previously for this?

[kkovaacs]

Looks good to me!