API improvements suggestions

We used miniscript to identify which M of N PubKeys correspond to the M signatures in a spent M of N multisig output, for the P2SH, P2WSH and P2TR (script path) outputs.

I wanted to make two suggestions about potential Miniscript API improvements:

1. The interpreter.verify_sig() function requires prevouts argument, that is a vector of all outpoints corresponding to all the transaction’s inputs. Only in the case of Taproot the signed message commits to the scriptPubKeys of all outputs spent by the transaction, per BIP 341. Even for the Witness case per BIP 143, only the amount from a single outpoint corresponding to the input being verified is required (and not the whole vector). For the legacy P2SH or bare script cases, no outpoints are required. Preparing this prevouts Vec for all legacy cases is an unnecessary computation in the program that is using the miniscript API.
   We suggest either:

   • making the prevouts argument an enum with three variants: ALL(& [T]) (for taproot), ONE(T) (for witness) or NONE (for legacy)
   • or replacing the interpreter.verify_sig function with three functions interpreter.verify_sig_taproot(), interpreter.verify_sig_witness(), interpreter.verify_sig_legacy().

2. interpreter.verify_sig() returns false instead of a concrete error. For example, we mistakenly passed a prevouts vector that did not correspond to inputs length. The Error::PrevoutsSize that cache.taproot_script_spend_signature_hash() returned was transformed into false here, and returned.
   It would be desirable to return a Result<bool, sighash::Error> from interpreter.verify_sig(), instead of bool.

Start with src/interpreter/mod.rs, especially interpreter.verify_sig() and the linked cache.taproot_script_spend_signature_hash() call. Compare the proposed prevouts API alternatives and Result<bool, sighash::Error> behavior across legacy, witness, and Taproot cases. Done means an agreed API design addresses both suggestions without converting concrete errors into false.