ENH: Categorical.unique can keep same dtype

[topper-123]

There are situations where we want to keep the same dtype as in the original after applying unique For example

>>> dtype = pd.categoricalDtype(['very good', 'good', 'neutral', 'bad', 'very bad'], ordered=True)
>>> cat = pd.Categorical(['good','good', 'bad', 'bad'], dtype=dtype)
>>> cat
[good, good, bad, bad]
Categories (5, object): [very good < good < neutral < bad < very bad]
>>> cat.unique().dtype == cat.dtype
False  # this is a bug IMO, but others may not agree

Even if it's not a bug, there are situations where we want the comparison above to return True. To alleviate the above, I've added a new parameter, so we can do

>>> cat.unique(remove_unused_categories=False).dtype == cat.dtype
True

Helps #18291, but does not close the issue.

[jbrockmendel]

id rather just change/fix the behavior than add a new keyword

[topper-123]

I´d be +1 on that, but I´ll wait on input from others. One consideration is that groupby with categoricals uses the current behaviour. Changing groupby behaviour without deprecation is not a good idea, IMO.

I thinik it would be reasonable to change the behaviour of Categorical.unique, but keep the current behaviour of groupbys. Then we can seperately discuss how the groupby behaviour can be properly deprecated and changed.

[jreback]

agreed this is very tricky and adding a parameter is just api bloat