Revise guidance to require a minimum upper bound for supported ranks

[kgryte]

This PR

• resolvesArray objects of arbitrary rank are infeasible - require a reasonable range of ranks instead #479 by revising the guidance requiring conforming implementations to support arrays of arbitrary rank and specifying that array libraries must support, at minimum, arrays of rank 4. Additionally, array libraries will be required to explicitly document their maximum supported rank.

Related

• RFC: add inspection support for querying the maximum number of supported dimensions #694

[kgryte]

cc @leofang@Zac-HD

[Zac-HD]

If array libraries have a consistent way to retrieve the maximum dimensionality they support, Hypothesis can provide more informative error messages. iirc Numpy has a MAX_DIMS constant somewhere...

[kgryte]

@Zac-HD Indeed, that is part of the rationale behind #694. Once we merge the inspection API in, we'll add support for querying the maximum number of allowed dimensions.

[rgommers]

This LGTM, thanks @kgryte. I believe this aligns with previous discussions, and won't be a problem for any implementation (while 4 was the default maxdims supported by an implementation based on Kokkos IIRC).

[rgommers]

This is pretty straightforward. I'll aim to merge this in a few days unless an objection appears.

[kgryte]

As this PR has received multiple approvals and no objections, will go ahead and merge. Any further changes can be addressed in follow-up PRs...