Refactor la.Vector.norm

[nate-sime]

Closes #3103.

Refactor la.Vector.norm into a function to reflect the C++ layer.
Add implementation of integer types for la.Vector.
Redesign the la.Vector tests to parametrize over dtype as well as xfail on unsupported (or not yet implemented) norm types.

[jorgensd]

Looks good to me, even if it means that we change the API:P

Only request is to add support for l1 and l-inf norms for integers, as they are supported. "l2" is clearly not supported.

I guess if @garth-wells or @IgorBaratta or @chrisrichardson has any opinions on this please chip in:)

[nate-sime]

Regarding supporting norms of vectors of integral types, I'm having a little trouble coming up with a neat way to deduce the underlying type of a vector of complex whilst also handling those integral types. The outline for this is clearly in dolfinx::scalar_value_type_t; however, this would either require another similar definition with carefully chosen name, or expansion of concept scalar which I don't believe fits in with the scientific computing philosophy. Perhaps this could be approached in a future pull request?

[garth-wells]

I have some comments - will post tomorrow.

[garth-wells]

Looks good - just a micro change suggestion.

[nate-sime]

I think this one's ready to go now if we can put it in the merge queue, @garth-wells ?