Add support for 64-bit indices and large arrays

[gonzalobg]

This PR adds support for 64-bit indices and large arrays. It is blocked on merging #186 .
Tested the following up to 141 GB of memory:

• cuda
• thrust
• std-data/indices/ranges
• omp
• raja
• tbb
• ocl

The following fail, and I need to minimize and report a bug; this PR does not regress them for small sizes:

• kokkos
• acc

The following is not tested:

• hip: @tom91136 it would be good to test it and test whether intptr_t or size_t is better for the indices (or no difference).
• sycl

[gonzalobg]

@tomdeakin @tom91136 This is blocked on #186, but is rebased on it now, and I've fixed the last thing that TomD requested, so should be ready once CI is green.

[tomdeakin]

@gonzalobg - think this is good to merge now too, right?

[gonzalobg]

Yes, I've rebased this on top of #186, and all tests also pass :)