Figure out rebroadcasting / RBF strategy

[tnull]

We should track on-chain transactions and (auto) rebroadcast/RBF/CPFP them periodically until they confirm.

[tnull]

In order to accommodate unreliable network connections, we probably should cache any failed broadcasts and only drop them after they have been retried a couple of times (with exponential back-off).

[tnull]

This was marked completed even though not fully addressed. Reopening.