Dealii: (Re)implement restriction/prolongation/interpolation matrices in FE_Nedelec/FE_NedelecSZ

[tamiko]

The current code quality in FE_Nedelec is very poor. The setup of above matrices/functionality is riddled with bugs (#326, #328, #332, ...) and due to missing caching FE_Nedelec::convert_... is incredibly inefficient.

While at it, we should replace the bulk of the implementation with #NedelecSZ.

In reference to #4992

[tamiko]

@dealii/dealii @rosskynch My suggestion for 9.1 still stands: Reimplement all missing features in FE_NedelecSZ and rename it to FE_Nedelec while removing the old code entirely.

[drwells]

I would still like to do this (based on the list of issues with our current nedelec implementation). Can we still fit this in the release @tamiko?

[kronbichler]

Talking for @tamiko now, I can review it today (if it is not too late here in Europe) or tomorrow morning. I guess it depends who decides when to make the branches.

[tamiko]

@drwells I would very much like to branch out as soon as everything we want to have in the release is in. How complete is your PR (will you get it in by Monday late)? :-)

[drwells]

I hope you are joking! I know nothing about the Nedelec element and did not mean to imply that I would volunteer. I apologize for the confusion.

[tamiko]

Ah, IC. Postponing to 9.2 then. This is at least a 1 - 2 week project.

[masterleinad]

It seems there was no action on this one from anyone and this takes more time. Postponing again.