rpcdaemon: eth_getTransactionReceipt performance degradation at high rate

[lupin012]

Looking at these performance measurements, it is possible to spot some degradation when increasing the request rate (i.e. query-per-second/duration, QPS/T):

• at rate 10000:20 we have max latency about 15 ms in Silkworm, a few seconds in Erigon
• when rate becomes 20000:20 in both systems max latency is over 20 seconds

Hence, Silkworm is showing quite a large performance degradation when QPS goes from 10000 to 20000.

In PR #1958 we tried to offloading some processing (specifically, the main hotspot highlighted by perf in laptop machine) to worker threads, but performance-wise things get even worse.