make interface topologies their own class

[gertjanvanzwieten]

Currently all topologies define both transforms and opposites. If a topology is not an interface then opposites are set to equal transforms. In this situation it would be better to not define them at all so that evaluation of function.opposite fails, pointing out a mistake in the formulation.
Suggestion: let's remove 'opposites' from Topology, and introduce a new InterfaceTopology that depends on a reference sequence and two transform sequences. It will lack connectivity, which is already the case for most interfaces and should not be an issue in typical applications. The StructuredTopology may still supply its own (structured) boundaries with opposites.