Document importing should a minimal amount of work if the document already exists

This is sort-of the document equivalent of https://github.com/n0-computer/iroh/issues/1947. If RPC calls to import the same document are made in quick succession (as happens with https://github.com/expenses/usd-iroh-resolver), the RPC calls will eventually stall out (unless a timeout is used) and the iroh node will produce a bunch of sync and gossip warnings:

2024-02-02T03:44:28.897392Z  WARN gossip{me=5edsipymvbm2kl2y}: iroh_gossip::net: dial failed: failed connecting to provider peer=PublicKey(7siesmyj2xdrlsti)
2024-02-02T03:44:28.897519Z  WARN sync{me=5edsipymvbm2kl2y}: iroh::sync_engine::live: sync failed origin=Connect(DirectJoin) err=Failed to establish connection

Caused by:
    0: failed connecting to provider
    1: timed out

I managed to work around whatever issue is happening with iroh by checking if the document to import exists locally first:

if let Some(doc) = iroh.docs.open(ticket.capability.id()).await? {
    return Ok(doc);
}

iroh.docs.import(ticket.clone()).await

This will not produce the expected behaviour if the ticket contains the addresses of nodes that aren't currently being synced with, as these will not be sent to the iroh node.

Ideally iroh.docs.import should perform the same actions as iroh.docs.open if the document already exists and there are no new node addresses, and if there are new nodes, should join these in the background while returning the document the same was as iroh.docs.open.

Start by comparing the behavior of iroh.docs.open and iroh.docs.import, using the repeated-import scenario and iroh issue 1947 for context. Check how existing documents and ticket node addresses are handled. Done means repeated imports return promptly like open while still incorporating new node addresses in the background.