fix(rpc/eth): remove cache when reorg occured

[jsvisa]

What's the problem?

eth_getBlockByHash returns the ReOrged block;

How to reproduce

We're running ETL process to fetch blocks and transactions from the RPC server, and then save them into our DB, and after some epoch(eg: 64 blocks), recheck the saved blocks to see if it's forked blocks or canonical ones.

If it's a forked one(by checking the block hash can continue to be queried to determine), we need to remove the saved transactions and re-fetch the block with the same block number.

And this process runs when in geth's RPC, but failed on reth. After some digging, found geth will remove the hash -> number mapping when reorg occurs: https://github.com/ethereum/go-ethereum/blob/df3f0a81a7ef6fd56c90675c32b6d732380a9fd2/core/blockchain.go#L1322-L1327

Reth will also remove the mapping in the underlying DB(by restarting reth, this issue disappeared), but the cache is not fully removed, so I propose this PR.

[onbjerg]

makes sense to me, pending @mattsse

[mattsse]

oh very nice find, ty.

this fix is fine because we also reinserting new blocks into the cache again:

reth/crates/rpc/rpc-eth-types/src/cache/mod.rs

Line 340 in d395df2

self.full_block_cache.insert(block_hash, block);