Add error returns to maxGridDiskSize

[isaacbrodsky]

This PR also changes the signature for this function to support k values where the number of cells is more than an int32_t can represent (i.e. k=26755, maxGridDiskSize(26755)=2147570341)

[coveralls]

Coverage increased (+0.01%) to 98.185% when pulling f17615e on isaacbrodsky:error-returns-max-grid-disk-size into efad372 on uber:master.

[dfellis]

Should the benchmarks check for the error condition?

[isaacbrodsky]

Updated for the setup part. I didn't feel the benchmark should check this as part of what it times since this isn't part of the code we're intending to measure.

[dfellis]

Only a minor comment on the validity of benchmarks that all fallible functions but assume correctness.

[nrabinowitz]

Would it make sense to check somewhere that this value is less than the max number of cells at the current res? E.g. should kRing(res0cell, 100) throw E_DOMAIN?

[isaacbrodsky]

I have a concern that in some cases this will overflow, in that case we should consider returning the maximum number of cells?

[nrabinowitz]

Yeah, that's probably better than returning an error (and should be correct, in that I think k-rings covering more than the total number of cells should return all cells at that res, right?).

We should be able to check a constant to avoid integer overflow, right?

[ajfriend]

One minor comment: Should we standardize the type we use for distance? Here, it looks like we're using int, but we use int64_t in gridDistance.

h3/src/h3lib/include/h3api.h.in

Lines 701 to 702 in cd47015

	DECLSPEC H3Error H3_EXPORT(gridDistance)(H3Index origin, H3Index h3,
	int64_t *distance);