* Always use simple LIST syntax if it succeeded once after a failed
complex query
* Reject asking for a LIST twice
* Quickly discard any ISUPPORT messages that don't contain things we
need to use
* Detect the server name and drop any messages from the server
Fix a bug where the channel allocation algorithm failed due to
necessary relays not having been provisioned yet.
Passed the newly created relay numbers to the allocation function
and assumed their CHANMAX would be the same as all other relays for
the same network.
* Allocate channels to relays only if they have free space based on
their chanlimit value
* Minify channels by removing ones that are already covered before
passing them off to be joined
* Low-key channel joining with incrementally increasing delay
* Spin up needed instances to be able to cover a certain channel space
* Fix provisioning functions to prevent race conditions with lots of
relays being created at once
* Tweakable switchover from covering all channels to only covering
channels with more users than the mean of the cumulative user count
Mark Veidemanis