Add support for 64-bit `CeedInt`

[zatkins-dev]

For sufficiently large problems, the indices on a single process may exceed the $2^{31}-1$ (~2 billion) index limit.
It would be good to have a compile time option to enable 64-bit integers.

This would also improve interoperability with pointers from other sources, e.g. PETSc.

[jrwrigh]

When would that happen? The way HONEE handles it, we use CeedInt for partition-local indexing only and PetscInt for global indexing. Global indexing can absolutely exceed 2 billion, but I'm not sure when local indexing ever could? Have you run into this limitation?

[zatkins-dev]

I haven't run into this issue, nor do I think it would happen commonly. The most likely case would be with at-points operators, since they effectively require one index per quadrature point. Admittedly, I still think it's unlikely, but I don't see a big reason to not support it

[jrwrigh]

I don't see a big reason to not support it

I think the big thing is potential confusion to users. IMO, 64-bit CeedInt is a code smell that has a non-negligible performance impact.

If we choose not to support it, I do think it'd be worthwhile having a blurb somewhere in the documentation for why we don't have the option. Or if we do support it, why it's probably a bad idea.