Decoding the encoding of a valid instruction yields the original instruction.
That is, decoding is left inverse of encoding, over valid instructions.
Theorem:
(defthm decode-of-encode (implies (instr-validp instr feat) (equal (decode (encode instr feat) feat) (instr-fix instr))))