feat: support different number of effects for each qubit

[timmintam]

Extend the interface of the RandomizedProjectiveMeasurements class to support different number of effects for each qubit.
We may need to move away from np.ndarray input types for bias and angles to sequences of sequences.

[timmintam]

As a consequence of #94, the argument bias is now allowed to have some values set to 0 for RandomizedProjectiveMeasurements instances. In practice, this changes the number of effective effects for each qubit.
Although it not the cleanest implementation (one still has to enter some dummy values for angles even if the corresponding bias is set to 0) it works. Maybe we will want to have a cleaner implementation in the future.