The latest rust-miniscript version deprecated the helper we were using,
in favour of one that gives the maximum size difference of a transaction
input before and after satisfaction. The new helper differs in that it
does not account for the empty ScriptSig byte (which uncovered we were
actually double-counting it), and assumes the non-satisfied transaction
input is already part of a Segwit transaction (which we rectified).
This uncovered a mistake in the computation of the witness script size
in the unit test. We also get rid of the needless wu_to_vb() standalone
function.
b9753b48d0