We mine quite a few blocks in this test to trigger a htlc timeout,
so it can be pretty slow. This fix adds assertions for additional
"state steps" that happen in between the failing of the htlc on-chain
and asserting that we're fully cleared out so that slower running
machines won't timeout.
This itest adds a test that we still propagate blinded errors back
properly after a restart with an on-chain resolution. The test also
updates our sendpayment timeout to longer so that there's time to
resolve the on chain claim.
We don't support receiving blinded in this PR - just intercept and
settle instead. The HTLC's arrival on the interceptor indicates that
it was successfully forwarded on a blinded hop.
