simplex support comments

[tjhei]

After @peterrum 's presentation yesterday, I have a few comments worth discussing. I would propose that we have an API round-table via video chat at some point soon. My notes so that I don't forget:

• I would think cell->cell_type() might be neater than cell->reference_cell_type() (Rename variables named 'reference_cell_type' to 'reference_cell'. #11688)
• Should we deprecate static info in GeometryInfo? Con: a lot of changes for users. Pro: consistency, bug finding.
• Should we consider deprecating the default mapping arguments? This needs a longer conversation, but I think I am in favor.
• Change default mapping to something that is Q1 or P1 depending on the cells used and call is something like MappingLinear (I know that it is not strictly correct)?
• Detect functions inside the library that rely on quads and either fix or add an Assert(tria->is_quads_only()). This would involve (temporarily) removing GeometryInfo, MappingQ, etc..

[bangerth]

Here's another thing I'd like to see: In classes like QGauss and QTrapez, the Q doesn't actually indicate a quad/hex style quadrature, but that it's a quadrature.

So I don't think we should introduce PGauss and PTrapez. Rather, I'd suggest giving the constructor an argument of type ReferenceCellType (or whatever that should be) and by that either let it construct a tensor product rule or use whatever is the appropriate choice for that reference cell kind.

[bangerth]

I can't seem to find any evidence of the PTrapez name mentioned above and assume that I just never looked it up.

[drwells]

I think we have addressed all of these problems in various ways - we have ReferenceCell, many functions assert that the mesh is all hypercube elements, #10310, etc.