eth: fix deadlock casued by multiple subscriptions/requests

[nick1udwig]

Problem

Our default eth provider, providerfren.os, frequently hangs, requiring restart.

Solution

Fix deadlock by minimizing the amount of time locks are acquired for. In particular move everything that communicates with a timeout outside of a lock.

Testing

To test I:

1. Set up an eth provider using this binary or master
2. Modified kns_indexer to infinitely loop sending get_logs(), by commenting out

   kinode/kinode/packages/kns_indexer/kns_indexer/src/lib.rs

   Line 163 in fbf961e

   break;

3. Run one node using looping kns indexer, pointed at eth provider
4. Start a second one & stop if it successfully connects, repeating until either witness deadlock or get bored

Using that protocol, I get a deadlock immediately on master, whereas I have not witnessed a deadlock on this branch.

Docs Update

None

Notes

None

[dr-frmr]

amazing