Allow sequence to accept stacked np arrays if the feature space is Box

[jjshoots]

Description

This allows the Sequence space to accept numpy arrays of size [n, *base_size] if the feature space of the Sequence is Box(shape=base_size).

Fixes # (issue)

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

[pseudo-rnd-thoughts]

Personally, I'm very uncertain of this implementation.

I don't understand the difference between the new and old contains functions. Could you provide an example where the old would fail and it won't with new?

There are a couple of places in the code that need modifying if we want the whole space to support stacked sequences

[jjshoots]

@pseudo-rnd-thoughts The simplest case where the old would fail is the test_stacked_box test.

We're not really accepting stacked sequences, we're just allowing the sequence space to accept np.stack([array_a, array_b, array_c]) if the base feature space is Box.

More concisely, this is the only difference:

import numpy as np
>>> sample = np.float32(np.random.rand(5, 3))
>>> sample
array([[0.55573964, 0.5059766 , 0.1776721 ],
       [0.78431445, 0.19933775, 0.75381595],
       [0.82225245, 0.5078049 , 0.24258688],
       [0.66441643, 0.7965829 , 0.9005535 ],
       [0.31705654, 0.6865207 , 0.25575027]], dtype=float32)
import collections.abc
>>> isinstance(sample, collections.abc.Iterable)
True
>>> isinstance(sample, collections.abc.Sequence)
False

[pseudo-rnd-thoughts]

Could the contains function not just have Sequence changed to Iterable.
Personally, I prefer the contains functions to check if the sample is positively contained within the space rather than negatively if that makes any sense.

I guess we can add this but @kad99kev will be looking into adding this feature to the whole space.