This adds five test vectors for BIP352 cases the file does not cover; the changelog entry names them. reference.py computes every expected value and passes on the whole file, and btclib-wallet, which I maintain, agrees on all five (run). The version is 1.1.4 because #2323 takes 1.1.3; I will rebase whichever lands second.
Was I right to remove three cases that cannot be consensus-valid, under the scope #2339 proposes? They are a P2PKH input whose pubkey hash matches 33 bytes that are not a valid key, a P2TR input with an empty witness, and a P2TR input whose output key is not on the curve. reference.py keeps code for all three, and no vector reaches it now. I kept inputs whose only flaw is a placeholder signature: the file has no spending transactions, and the ECDSA signatures these vectors reuse from it have no sighash byte.