Feature request: witness size estimator that understands key availability

[apoelstra]

When estimating witness size we should understand which keys (or hash preimages, timelocks, etc) are definitely available, definitely unavailable, or 3rd-party controlled.

[apoelstra]

For context - this is something that @sipa, @sanket1729 and I thought we had a simple and efficient algorithm for, but we now believe we were wrong, and don't have a good idea of how to proceed. Input from algorithms experts would be appreciated.

[sanket1729]

This is being addressed in #481. Really cool that long time issues are being addressed :)

[apoelstra]

I think to the extent this issue makes sense, it was addressed by #592 :).