Retry locking Monero

[binarybaron]

If Alice fails to lock her Monero on the first try, she will not try again.

There are a number of reasons the locking could fail:

• Insufficient funds due to multiple swaps being started concurrently. One of which locks all the funds before the second one gets to.
• Network errors (no connection to daemon, is common during syncing)

We should retry continously until the cancel timelock expires.

https://github.com/UnstoppableSwap/core/blob/4867d2713f5cdcc2aa282760ad88c12b17daf487/swap/src/protocol/alice/swap.rs#L113-L132

[louislaugier]

We run an ASB and kept seeing swaps get stuck as "xmr is refundable". The root cause is the concurrent locking described here: when two swaps enter the Monero lock phase at the same time they can select the same output. wallet2 only marks an output spent once its lock transaction is relayed, so the second lock transaction is a permanent double-spend that monerod rejects, and that swap hangs until the taker refunds.

I opened #1193 to fix it. Instead of retrying after a failed lock, it serializes the construct-to-first-publish window with a process-wide mutex, so at most one unpublished lock transaction exists at a time. Verified with the concurrent_bobs_before_xmr_lock_proof_sent, concurrent_bobs_after_xmr_lock_proof_sent and happy_path integration tests.

This looks like the same concurrent-lock problem. Worth reopening?