wallet: failed txs due to dropping connection during long operations

When doing some operation that takes a long time (10s of minutes), such as a sweep_all with many outputs (i think ive seen this also for get output distribution), after it finishes constructing the txs, it fails to broadcast them, citing no_connection_to_daemon.. followed by an immediate reconnection. The txs dont broadcast, so you're left with wasted time and forced to manually construct in parts.

This happens using cli and rpc. daemon specified on the command line (not via rpc command). Using local daemon, trusted

The fix here is likely just to reconnect before the send operation.. or some way to ensure the connection is maintained in the first place.