Make process_onion_failure always blame an SCID for non-payment-failed

[TheBlueMatt]

We should try hard to make process_onion_failure always return a blamed SCID if we aren't considering the payment as retryable. If my reading is correct, the only case that needs to be covered to accomplish that is the case where we do not manage to read a valid failure packet at any hop. Its not clear to me what the right answer is in that case, though - can we determine who screwed up the crypto in some way?

Once we do that, as a followup, we should make Event::ProbeFailed::short_channel_id an non-Optional field, which will need to happen in the same release as its original inclusion, making this a blocker for 0.0.110. See #1567 (comment)

[TheBlueMatt]

Don't think there's anything we can do here, immediately, even though we very much should do something. Probably blocked on #1625 or some other mechanism where we can assign a penalty to all the nodes in a path without setting their available capacity lower, which isn't really the right thing to do for an entire path.

[TheBlueMatt]

I think we should just wait for attributable errors at this point.