BIP375: check SegWit versions in non-witness UTXOs

[zjubiology]

BIP375 requires constructors and signers to reject PSBTs that combine silent payment outputs with inputs spending SegWit v2+ outputs.

The reference validator currently performs this check only when the prevout is provided through PSBT_IN_WITNESS_UTXO. BIP174 also permits SegWit inputs to provide the full previous transaction through PSBT_IN_NON_WITNESS_UTXO, so an input containing only that field can currently bypass the SegWit-version check.

Reuse the existing _script_pubkey_from_psbt_input() helper when validating input eligibility. This extracts the referenced scriptPubKey from either UTXO field.

Also add an invalid test vector containing:

• a silent payment output;
• a SegWit v2 prevout supplied only through PSBT_IN_NON_WITNESS_UTXO;
• a matching previous transaction ID and output index.

[zjubiology]

@achow101 @andrewtoth Could you please take a look when you have a chance? Thanks!

[fametrano]

The validator change holds: at ea1192c, master (927b6de) accepts the new vector under input_eligibility, this branch rejects it with "Input 0 uses segwit version > 1 with silent payments", and all 43 vectors pass.

The vector's PSBT_IN_PREVIOUS_TXID is byte-reversed. The field holds 9630266b…0c73c9; the double-SHA256 of the PSBT_IN_NON_WITNESS_UTXO transaction is c9730c1b…263096. BIP 174 requires the two to match ("its hash must match the hash specified in the prevout"), and the six existing vectors with a non-witness UTXO store the hash as computed, so a parser that checks this rejects the PSBT for the wrong reason. Reversing the 32 bytes of PSBT_IN_PREVIOUS_TXID, and of prevout_txid in supplementary to match the other vectors, fixes it.

[fametrano]

ACK 242bfbe

The new vector's PSBT_IN_PREVIOUS_TXID now matches its non-witness UTXO, the test runner passes 43/43, and the vector is still rejected for the segwit version.

[zjubiology]

ACK 242bfbe

The new vector's PSBT_IN_PREVIOUS_TXID now matches its non-witness UTXO, the test runner passes 43/43, and the vector is still rejected for the segwit version.

Thanks for the review and ACK!

[zjubiology]

@fametrano Thanks for your review! If there's anything need I to do, please feel free to let me know.