feat: include units for duration and angles

[qartik]

Another change is that "duration" is now an optional field of MOp

Closes #6

[qciaran]

The solution is a little different then what I was thinking. It looks like pure floats are kept as pure floats and assumed to be radians and you have the option to include a "pi" unit; while, for durations it looks like you always have to specify the unit.

To be consistent between angles and durations, I was thinking more, if the user can choose to supply both angles and durations as purely floats and that the Model will recast it as [float, "rad"], [float, "s"], respectively. (That is, float == [float, "rad"], float == [float, "s"], respectively.) The user/generator of the JSON can also choose to be explicit with typing.

Is that a pain, e.g., for validation? Or do you think there are other issues with that? I kind of like your solution actually. The only thing that causes me pause is the inconsistency (but maybe they don't have to be consistent).

[qartik]

@qciaran I won't worry about consistency between angles and durations. For angles, as you argued, it makes sense to have a default as it's both obvious and widely accepted that without any units, angles are radians.

For durations, however, it's quite perplexing to not see any units attached while reading a number.

As a further improvement, I am considering the unit of angles to be mentioned only once in the QOp, so perhaps a field such as angles_unit that has the default value "rad" (or None representing radians) and other acceptable values like "pi" or even "deg". To me, it does not make sense to provide angles in multiple units for a single gate. Moreover, providing units with each angle makes it unnecessarily verbose. Do you foresee that being problematic?

For example, compare:

QOp(metadata=None, qop='PhasedX', returns=None, args=[('q', 0)],
    angles=[(0.5, 'pi'), (0.5, 'pi')]),

vs

QOp(metadata=None, qop='PhasedX', returns=None, args=[('q', 0)],
    angles=[0.5, 0.5], angles_unit="pi"),

[qciaran]

Since radian are usually given as unitless, that is fine. However, I would go with "rad" instead of None to be explicit. (I don't think we need to bother with "deg" since that isn't used in any scientific computing that I am aware of.) Sure consolidating the angles_unit to a separate entry is fine, although keeping them with the angle does not disturb me either. If you do go with a separate field for the unit, I think a consistent approach should usually be taken so you don't have to remember special cases. Therefore, if you go with angles_unit you should also have duration_unit.

[qartik]

Agree with everything, except forcing duration_unit as a separate field when the type Duration is sufficient to capture a duration fully. Perhaps a compromise might be better?

How about the following definition for angles?

angles: tuple[list[float], Literal("rad", "pi")] | None = None

[qciaran]

Yeah, that works for me.

[qartik]

@qciaran ready for your review. Thanks!

[qciaran]

Looks great!