soroban-rpc: Audit and analyze retries on any external calls

[paulbellamy]

What problem does your feature solve?

Currently, there is not much retry-handling around external calls (e.g. history archive service, or filling the db)

What would you like to see?

We should go through everywhere we do an external operation which could fail, and write a comment with the retry policy for that (then implement it).

For instance, at startup, wait for the first checkpoint if need be (otherwise soroban-rpc fails right away)

What alternatives are there?