bv_simplify should always produce equalities on Z, but due to a bug it turned bv_wrap n (bv_unsigned b) = bv_wrap n (bv_unsigned b) into b = b, which means that bv_solve fails on bv_wrap n (bv_unsigned b) = bv_wrap n (bv_unsigned b). This MR fixes this bug which is caused by apply applying an ↔ lemma in the wrong direction.
