core.util.validate_regular_sampling() appears to contain redundant call to np.unique()

[ceball]

https://github.com/ioam/holoviews/blob/master/holoviews/core/util.py#L1653

I happened to notice that creating an Image becomes surprisingly slow as one dimension gets bigger (for a fixed overall Image size). That led me to notice that validate_regular_sampling() contains a call to np.unique() (which sorts). I'm not clear on the purpose of the call to np.unique().

I could demonstrate the problem I'm talking about, and fix it, but first I'm opening an issue in case there's something I'm missing about the need for unique().

[jbednar]

I don't quite get what the unique achieves here either, but if it's just about squeezing repeats rather than true uniqueness, you could do something that doesn't sort, like:

[philippjfr]

I think there is indeed no need for np.unique and this could simply be:

diffs = np.diff(values)
return abs(diffs.min()-diffs.max()) < abs(diffs.min()*rtol)

[philippjfr]

This was fixed in #2293