This is kinda like how matrix diagonalization can be used to solve any problem which is expressible as a linear system of equations, and to some degree any continuous function can be approximated by a linear system, so a BLAS + LAPACK accelerator is a "universal simulation engine."

You could probably build a generic Shor's algorithm quantum computer that is able to both factor integers and break elliptic curve keys. But the same quantum computer wouldn't be usable for Grover's algorithm to find the presage of a cryptographic hash. This is what I mean by there not being a "universal quantum computer" in the same way a Turing machine is a universal classical computer. Quantum computers by their very nature are ASIC implementations of specific algorithms, even though those algorithms might have some multi-domain applicability.