Allow bypassing of coupling vertices detection in the adapter

[IshaanDesai]

Currently the adapter requires a filter function which it can apply on a mesh to detect the vertices on the coupling boundary or volume. While this API is in alignment with the internal functionality in DOLFINx, it severely limits the users ability to define coupling boundaries or volumes of arbitrary shapes. The adapter could simply accept vertices that the user provides it, instead of doing the filtering itself. While this may seem as though functionality of the adapter is being turned into a burden for the user, it will increase flexibility of the adapter.

This issue is somewhat related to the issue of being able to define the coupling domain with mesh tags #8

[steffenger]

I think this is part of a bigger discussion, namely what functionalities should be handled by the adapter and what functionalities should be outsourced to the user at the end of the day. From my view, a adapter is a bridge between the solver and preCICE, which improves the user experience compared to the coupling between the solver and preCICE directly. And this is really tricky since fenicsx is probably different to other solvers, because it is also an API and it comes with a lot of functionality related to e.g. creating different meshes, function spaces or doing interpolation. Therefore just providing the right vertices to preCICE and read/write them could be enough to handle a lot of coupling cases (e.g. also higher-order spaces). So the question is what is the central building block of functionality, which should be done by the adapter? Since the adapter is also based on the fenics adapter, we also need to check what has changed between fenicsx and fenics in terms of functionality or name convention, e.g. they changed from Expressions to lambda functions.