Update monitors for new chain data without blocking the world

[TheBlueMatt]

We currently hold a full lock on a full ChainMonitor when connecting blocks for all our ChannelMonitors. We shouldn't do that, we should drop the lock after each monitor and give users a chance to use their node in between full monitor syncs. Should improve #1528 a good bit.

[TheBlueMatt]

Moving back to 117 cause this is causing substantial user pain.

[G8XSU]

Current approach : Acquire read lock on all monitors, read keys. Acquire write lock only per iteration.

Concern: If we don't have write lock on all monitors,
Can there be an issue that some monitor got added to the map while we were acquiring write lock only for one monitor at a time?
And hence that newly added monitor got skipped for block_update.

[TheBlueMatt]

Yep, we'd have to track that and handle them specially at the end.