Add option to coerce all `ZZPhase` angles to [-pi/2, pi/2) when submitting

[cqc-alec]

[PabloAndresCQ]

Adding for context:

pytket-quantinuum outputs QASM/QIR circuits with angles in the interval [0, 4pi) but the machine actually runs gates with angles [-pi/2, pi/2]. Importantly, the way negative angles are run in the machine is different from positive angles, and this is something that Charlie's team would like to do benchmarking on.

So, there's two questions:

• Should pytket-quantinuum, and all other ways of submitting circuits to the Quantinuum devices guarantee that the angles are in the [-pi/2, pi/2] interval?
• Can the US side guarantee that a circuit in the {Rz, Rxy, Rzz} gateset with angles in the [-pi/2, pi/2] interval would not have their angles changed?

[PabloAndresCQ]

Here's a summary of what I've found out so far:

Note: to attempt to avoid confusion, all discussion of angles here are in terms of radians, rather than pytket's "half turns". I use Rzz to refer to the ZZPhase gate.

• When a QASM circuit is submitted to the US compiler with {compiler_options: {"parametrized_zz": True, "nativetq": "RZZ"}}, angles in the range [-pi/2, pi/2] for Rzz gates are actually run as such in the machine.
• This does not hold true for Rxy (aka PhasedX).
• pytket-quantinuum outputs QASM with angles in the range [0, 4pi).

To satisfy Charlie's requirement, we'd need to:

1. Apply a pass in pytket-quantinuum (after rebasing to the native gateset) that guarantees angles of Rzz gates are in the [-pi/2, pi/2] interval. This will require using the fact that Rzz(pi) = Z \otimes Z so that, for instance, Rzz(5pi/4) becomes Rzz(pi/4)*(Z \otimes Z).
2. Make sure that the appropriate compiler options are set when the user requests the option to coerce to the [-pi/2, pi/2] interval.