Improved API reference structure for probnum.problems

[JonathanWenger]

In a Nutshell

Changes the API reference structure for the probnum.problems and probnum.utils subpackage to make the hierarchy immediately understandable and closer to the actual package hierarchy.

Detailed Description

Improves and stratifies the sidebar menu for probnum.problems in the API reference, like so:

Related Issues

This implements #450 for probnum.problems.

[mmahsereci]

Thanks for doing these tedious bits. I like the re-arrangement of the problems docs. Just one comment on where the linalg utils should go.

[mmahsereci]

Souldn't these go under linalg API? I.e., probnum.linalg.utils instead of probnum.utils.linalg? It does't seem right to move sth specific to a general package.

[JonathanWenger]

Possibly yes. This should be a separate refactoring issue however. Since this is up in the air, I removed the utils part of the refactoring.

[mmahsereci]

Sure, we can discuss that later as well. It might be easier then to leave the utils part in this PR since it is easy to see which util function belong to linalg. and we can still see later if we want to move it. Does that make sense?

[pnkraemer]

The idea behind moving the Cholesky updates to (the thereby created) utils.linalg instead of linalg.utils was (a) Cholesky updates are a linear algebra tool for PN algorithms, and not a probabilistic linear solver, and (b) they are needed in e.g. statespace, and a linalg dependency for this small thing would've been a bit weird in my opinion. If you have stronger reasons to move it there, though, I am happy to discuss :)

[mmahsereci]

I see, then simply the namespace is confusing as is may imply that the utils.linalg are only required by linalg subpackage. If the utils are used by multiple subpackages they should stay where they are (perhaps with a different namespace.) Thanks for the explanation @pnkraemer !