Replace all results objects in 0.2 with arrays

[andrea-pasquale]

Originally posted by @alecandido in #896 (review)

[alecandido]

The dimension of the array will be:

(shots, sweeper1, sweeper2, ..., sweeperN)  # discriminated
(shots, sweeper1, sweeper2, ..., sweeperN, 2)  # IQ
(shots, sweeper1, sweeper2, ..., sweeperN, times)  # raw waveforms

without shots if an average is taken.

A function to yield the shape given ExecutionParametrs, list[Sweeper] will be provided (mainly for testing and validation, but also exposed to the user, documenting the meaning).

[alecandido]

@stavros11 @andrea-pasquale: I was considering having a fixed amount of dimensions, other than the sweeper ones. I.e. replacing "missing" dimensions with 1s.

This operation could be automatically undone with np.squeeze.

The idea behind would be to have a more predictable shape, and to keep the same rank (as much as possible).
On the other side, dimensions are useful: for the same reason for which we have a variable shape length because of sweepers, it is better to have a (143, 2) shape for a raw waveform than a (286,) one (is it raw major? column major?), and better no dimension at all if discriminated.

In principle, this may complicate the handling of different execution parameters (acquisition type and averaging mode). In practice, I expect that they have to be handled separately in any case...