Support "trivial encryption"

#1409 "moved final computations outside of the loop to be before the loop, so that the iter_arg of the loop can be a ciphertext type to match with the yielded ciphertext type."

In the future, we might get issues like this that can't be worked around by changing the computation.
In those cases, we could consider adding a `trivial encryption` operation that acts like a "cast" of the ptxt to a ctxt of a given order. IIRC, we can actually realize/lower this by just padding with zero polynomials., e.g., a plaintext polynomial p would become order-2 ctxt (0,p) decrypting to p - 0*s = p.