Prepare Plan::update_psbt_input for migration to rust-psbt

[nymius]

This is the simplest way I found to use the public Plan API in the Plan::update_psbt_input method so It can later be migrated to the rust-psbt project.

Notes to the reviewers:

• There is any reason to not expose the descriptor and template fields directly?
• There is a TODO in the code, but I'm lacking the context of it. It is something that should be fixed or a leftover?

[apoelstra]

• There is any reason to not expose the descriptor and template fields directly?

Nope, go for it.

• There is a TODO in the code, but I'm lacking the context of it. It is something that should be fixed or a leftover?

The context appears to be a few lines below where we do a loop for (pk, key_source) in data.key_origins which then builds a new tree in input.tap_key_origins. I think the TODO is that afillini imagined that he could somehow directly memcpy the tree structure over instead of doing a tree traversal?

Probably fine to delete the TODO. It dates to #592, like much of the "weird" stuff in this module, and likely wasn't reviewed since the new module was too big.

[nymius]

Thanks for clarifying _@apoelstra , I've removed the TODO as the functionality seems covered and the possible explanation is related to a performance improvement.

I've publicly exposed the descriptor field as it was the only field that I didn't have access to.

As removing the witness_template method would be a breaking change, and exposing the template field is duplicating the functionality provided by this method, I'm leaving as it.

I've copied the code from here into rust-psbt to verify it helps moving towards removing PSBT stuff from miniscript. You can check it in: https://git.rust-bitcoin.org/nymius/rust-psbt/compare/push-yormpqokyrxz...nymius/rust-psbt:push-xstopvrvruzp

[nymius]

I've removed the changes that weren't needed by rust-psbt to port update_psbt_input there.

[apoelstra]

ACK 0183420; successfully ran local tests