Support cache to retrieve ethereum logs

[crystalin]

Motivation

Currently the eth_getLogs is heavily used by some projects, sometime querying up to thousands of blocks in 1 request frequently (they could definitely improve it).
This has a huge impact on rpc nodes.

Suggested Solution

Because the blocks are constant, we should cache the ethereum logs so that requesting those would have a minimal cost.

I suggest to implement it directly there:
https://github.com/paritytech/frontier/blob/a9ec7ddd002b398937d5dae4d2b581a717b118b7/client/rpc/src/eth.rs#L352

Providing a new function like get_all_logs which would retrieve all the logs of a block and cache it.
This would allow to still use the cache even when there are filters.

Ideally, this on-the-fly cache would use an LRU eviction to ensure the most queries blocks would stay.

[crystalin]

Some tracing for the getLogs:

(You can open with speedscope.app , thread 27918 of
perf.getLogs.txt

[arthur-cw]

hey @crystalin, this is interesting, do you have any idea on how to do it?

[crystalin]

@arthur-cw This is already included in frontier
#540