Allow polyhedral methods returning subobjects to also return IncidenceMatrix

[lkastner]

Is your feature request related to a problem? Please describe.
It is sometimes hard to get at an IncidenceMatrix for certain polyhedral data, like cones in a fan, facets, etc. There are methods like ray_indices, but these are hard to find.

Describe the solution you'd like
Allow us to use the established principle of adding the desired return type. E.g. facets already allows to be called as facets(LinearHalfspace,...) and facets(Cone,...), why not add facets(IncidenceMatrix, ...)? Have a look at
491e3f3
for an example.

Describe alternatives you've considered
None, I want this.

Additional context
Usage of IncidenceMatrix should be extended, since it avoids many expensive operations.

The functions I had in mind are:

• maximal_cones(::Type{IncidenceMatrix},...)
• faces(...)
• cones(...)
• facets(...)
• maximal_polyhedra(...) and polyhedra if these exist. We probably need them for the PolyhedralComplex.

The Cone and PolyhedralFan case should be relative canonical. For the Polyhedron and PolyhedralComplex case there is a bit more to do to figure out what these indices mean. I am not a fan of splitting the IncidenceMatrix returned though, rather I'd think about introducing something like minimal_generators or something, returning an iterator with both RayVector and PointVector as they are in the VERTICES matrix. When someone works with the IncidenceMatrix I think we are allowed to assume that someone has a basic level of competence. Decomposing the IncidenceMatrix into a vertex and ray part would just make intersecting faces etc. much harder.

[YueRen]

Just wanted to add that this is not only applies to IncidenceMatrix, but also "normal" matrices.

For example, affine_inequality_matrix(facets(cube(3))) could be facets(QQMatrix,cube(3)).