Improve Topology construction and fix MPI/NBX bug

[garth-wells]

This change gives a mesh Topology a 'fully formed' constructor, i.e. a Topology object is complete from construction in that it has vertices and cells. Other entries are derived from these. This follows the DOLFINx design philosophy where we avoid creating half-baked objects with unclear state.

These changes triggered are seemingly unrelated bug in the NBX algorithm related to MPI. Consecutive, fast calls to the NBX could result in corrupted data since the non-blocking MPI calls used the same tag. The tag passed to MPI functions is now an optional argument.

Change also remove duplicated code in Topology constructors.

Resolves some of the point in #3506.

[jorgensd]

I would suggest refactoring the set_connectivity function as well, as well as set_index_map, as explained in #3506

[garth-wells]

I would suggest refactoring the set_connectivity function as well, as well as set_index_map, as explained in #3506

I think further changes should come later because they are much more intrusive for the interface, and some points need careful thought. This changes doesn't really affect the API (since users do not generally use the Topology constructor), and the PR fixes a bug.

[jorgensd]

I would suggest refactoring the set_connectivity function as well, as well as set_index_map, as explained in #3506

I think further changes should come later because they are much more intrusive for the interface, and some points need careful thought. This changes doesn't really affect the API (since users do not generally use the Topology constructor), and the PR fixes a bug.

What I am suggesting isn't very intrusive to the API as most users do not:

1. Attach index maps to the topology themself (as this is done through create_entities)
2. Does not clear the entities post creation.
3. Does set connectivity manually (as it is done through create connectivity).

The clear options would be a simple change for anyone using the current interface explicitly.

[chrisrichardson]

I think there is already enough in this PR.