allow cubical to accept larger arrays

[corybrunson]

The internal helper validate_arr_cub() ensures that a numeric array has dimensions that are acceptable to the cubical Ripser source code. However, these constraints in the source code ($2000 \times 1000$ for dim2, $512^3$ for dim3, and $64^4$ for dim4) seem arbitrary. It may be appropriate to warn the user before proceeding with the computation on a large array, but i believe these hard constraints should be lifted. This can be done easily in the dim2 case and with caution in the dim3 and dim4 cases.

@rrrlw does this seem reasonable to you? I'd also like to contact the authors to ask their reasons.

[rrrlw]

It does seem reasonable to me - as long as it works in a couple of basic unit tests and there is a warning message added for users, I don't see any issues adding

[corybrunson]

@rrrlw thanks. : )

I think the only challenge will be to allow the public/private "dense array" attributes of the classes to have dimensions determined from input rather than fixed. I'm reading up on how this might/should be done:

• https://stackoverflow.com/q/751878/
• https://stackoverflow.com/q/18234778/
• https://stackoverflow.com/q/23703837/