More tuples!

[glwagner]

Our function signatures are really long.

We need to use tuples / containers to pack arguments into Tuples or NamedTuples. Which groups make sense? The major ones I see are the source terms, the solution vector (u, v, w, T, S), and perhaps also parameters (eos, closure, constants, and g).

This may be helped by solving #163... but until then we can use tuples).

In addition @vchuravy suggested that we should pack the indices i, j, k into a tupled-index... perhaps I = (i, j, k) (or we can just use CartesianIndex).

For that, adding a convenience function to #249 to get the correct index associated with the intended ThreadsBlocksLayout for a given kernel might make sense.

[glwagner]

Not sure if this helps with some of the issues discussed in #290, but it might.

[glwagner]

Resolved bit-by-bit in many PRs.