Make on-close channel_update messages more robust

[TheBlueMatt]

When we force-close a channel we send an advisory channel_update to the currently connected peers to let the world know that the channel is now closed (by setting the disable bit). This doesn't do anything if we force-close on startup, because we have no peers. We should retry broadcasting these at some point in the future by caching them a bit. Other nodes currently dont do this (and absolutely should), but for now we should at least make our more reliable.

[shaavan]

Can I give this a try?

[TheBlueMatt]

Yes, feel free. Not quite sure how we should go about this, probably looking at channelmonitors on startup.

[shaavan]

I've tried expanding upon a different approach!

I've been monitoring our node's ability to reach out to anyone when it's time to broadcast, whether that happens at startup or any other timer while the node is running.

If, for some reason, we miss the broadcast, we'll just give it another go and resend the pending broadcast messages.

[TheBlueMatt]

We should still try to get this done in 119, but if it slips to 120, there's really no reason we need to delay the release for it.

[shaavan]

@TheBlueMatt
A gentle ping! :)

The PR resolving this issue (#2731) is ready for review.

[TheBlueMatt]

Important, but not worth holding up a bug fix release for.