Proper multi-material support for semidiscretize

[termi-official]

Currently semidiscretize does not have good support for subdomains with different physics. Needs to be fixed.

[termi-official]

For now I think it is better to go down a different road here.

First, we should add multi-material support at material-level, instead of at discretization level. The problem with having the support at discretization level is that it becomes hard to reason programmtically about the resulting function form of the semi-discretization, as no symbolic representation of the problem is available right now.

Second, for problems with different physics is crystalizes out that having different types for each instance is easier to manage than having a single struct which needs to be decomposed.