[Refact] Reviewing block generators

[Roland-djee]

Related to #266

All unitary operators have an hermitian generator such that
$$\hat{U}=exp(-i\hat{H})$$
In Qadence we have block.generator which is supposed to return $\hat{H}$. However, if we want to use HamEvo to then represent the original unitary, we have to specify a time value since it implements $exp(-i\hat{H}t)$. Currently, some blocks seems to have some inconsistent definitions.

This also affects GPSR, since there the paper assumes a generator structure of $exp(-iG/2x)$ for some gate $U(x)$.

Probable requirements:

• Make all gates follow the GPSR format, as suggested in [Refactoring] Enforce dimensionless parameter values passed to GPSR function #266
• Have a specific GPSR function to handle HamEvo differentiation, so that any arbitrary HamEvo(G, x) is processed in a way that the the $1/2$ factor is "considered" in $G$, and $x$ is an angle.

[Roland-djee]

@jpmoutinho You mentioned something about it in the planning, right ?

[jpmoutinho]

@Roland-djee yes, for @chMoussa to be aware of this while planning how GPSR is integrated into pyqtorch

[chMoussa]

Closing with #526