Consider adding an extra monitor update upon ChannelManager::funding_transaction_signed

[wpaulino]

@TheBlueMatt pointed out in #4336 a case where we end up with a force close when canceling a splice:

1. tx_complete is exchanged
2. Node 1 handles Event::FundingTransactionReadyForSigning, calls funding_transaction_signed, and sends its commitment_signed
3. Node 2 processes it, and sends its own commitment_signed and tx_signatures
4. Node 1 crashes without persisting the latest ChannelManager, and needs to handle Event::FundingTransactionReadyForSigning again
5. Node 1 chooses to cancel the splice this time around and sends tx_abort
6. Node 2 has already committed to the funding transaction via their tx_signatures, so they can't tx_abort back and force closes instead

Note that that is only worth fixing once we no longer force close due to a desync between the ChannelManager and ChannelMonitor persisted state.

[TheBlueMatt]

cc @valentinewallace and @joostjager After we fix things (via #4345 or otherwise) we'll need to do something like an empty ChannelMonitorUpdate to fix this case. It should be exceedingly rare either way though.