Add support of holes to contours generation

[sovrasov]

What does this PR do?

Adds support of representing non simply connected sets with contours: main contour describes the outer border of the set and "holes" are represented as child contours lying inside the space bounded by the outer one.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

[leoll2]

For self.shape and self.excluded_shapes, I would suggest to force the type to be np.ndarray or list[tuple[int, int]], but not both. Otherwise clients who consume contour.shape have to support both types, which isn't great.

Note: this is different from the class constructor parameter, which is instead an input interface. There, supporting more types is helpful for the clients. This principle is somehow formulated in the Postel law.

[sovrasov]

I personally prefer garbage in garbage out principle: the system should not help clients if they don't understand that they're doing. Looks like to much hustle for a wrong type annotation, which we can not just drop. Added dummy conversion to numpy.

[leoll2]

Thanks for fixing it!

the system should not help clients if they don't understand that they're doing

I don't fully understand this part; the type hints come from the library and help the client build robust code through static type checking (eg mypy). If the types are imprecise or loose, the client is forced to implement extra logic or suppress the warnings.