txscript: update ComputePkScript for a post taproot environment

[Roasbeef]

On the neutrino side, we started to use ComputePkScript to derive a pkScript based on the input witness of a txin. Post taproot, some of the assumptions we made are now incorrect, as an example: a valid taproot witness can be just a single element (the sig for a keypath spend).

On the neutrino side we've simply stopped using this function in the short term. In the medium term we should look into examning some of the new more concrete patters we can match un, as detailed by sipa in this comment: lightninglabs/neutrino#234 (comment)