[Quantum] Enfore QubitType representation by Linear Types

[lschuetze]

The value semantics of Qubits in the Quantum dialect enforeces the Linear Types. The missing piece is that qubits must be "consumed" by a quantum.deallocate(). On a side note, after a measurement the only Quantum operation that is allowed on the resulting qubit must be a quantum.deallocate().

[lschuetze]

On a side note, after a measurement the only Quantum operation that is allowed on the resulting qubit must be a quantum.deallocate()

If the hardware platform (or simulator) allows mid-execution measurements, then this does not hold.

[lschuetze]

The value semantics of Qubits in the Quantum dialect enforeces the Linear Types. The missing piece is that qubits must be "consumed" by a quantum.deallocate()

The current implementation only checks that a qubit is used at most once. It does not check if it is used exactly once. It is to be discussed whether we need the strong linearity or affine types (use at most once, i.e., 0 or 1) is good enough.