Updates necessary for SuiteSparse 7

[ViralBShah]

In SuiteSparse 6, SuiteSparse_long was deprecated, and you have to explicitly pick int32_t or int64_t. This makes things simpler, since we have a bit of complex logic in how we dispatch to cholmod_l routines.

We'll need to update the interfaces in line with this change for SuiteSparse 7 (we skipped 6). The vs/suitesparse7 branch in PR 48977 on julialang/julia is prepared for this and needs a new version of SparseArrays.jl that updates the interfaces. It passes on 64-bit platforms, and fails on 32-bit currently.

cc @fxcoudert @Wimmerer @Gnimuc

[PallHaraldsson]

Does it make a lot of sense to support SuiteSparse on 32-bit platforms, if it's a problem? We (at least I) want to get rid of it, for GPL-free Julia, by default, and could start that way for 32-bit Julia only? If people really need it they could use older Julia...

[I mean drop it, tough not necessarily SparseArrays too. For either or both it would be a precedence though...]

It's not clear to me 32-bit is even supported any more:

DrTimothyAldenDavis/SuiteSparse#221

msys2/MINGW-packages#14658 (comment)

https://github.com/DrTimothyAldenDavis/SuiteSparse/blob/42de2d841179f601ebbea83fbf8e7ef80c83039e/.github/workflows/build.yaml#L234

[rayegun]

We definitely cannot drop SuiteSparse for 32-bit. This would be a breaking change.

Tim is very receptive to stuff like this, and his solution sounds perfectly fine. Indeed it sounds like we might be able to get away with converting to 64-bit indices for certain arguments in the meantime.

[ViralBShah]

Will be accomplished in JuliaLang/julia#48977