Performance problem for some PI chains: explorer calls `eth_getLogs` with start block = 0 and end block = latest

[serejke]

We're running a Hyperlane PI on a custom EVM network (Axon ), whose RPC client is not capable of responding to eth_getLogs { startBlock = 0, endBlock = latest) quickly enough, and fails with a 504 gateway timeout. While we can request the performance improvement of the Axon clients, maybe a better solution would be if Hyperlane explorer could index logs by block ranges of 500-1000 blocks. With an all-blocks query, very popular Mailboxes will fail anyway (imagine 1M+ messages)