BUG: Fix segfault in Categorical.set_categories

[jschendel]

• closes Segmentation fault after set_categories() #24675
• tests added / passed
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

---

The fix basically amounts to changing self._codes --> cat._codes in the code block below:

pandas/pandas/core/arrays/categorical.py

Lines 850 to 855 in caf462c

	cat = self if inplace else self.copy()
	if rename:
	if (cat.dtype.categories is not None and
	len(new_dtype.categories) < len(cat.dtype.categories)):
	# remove all _codes which are larger and set to -1/NaN
	self._codes[self._codes >= len(new_dtype.categories)] = -1

Since the new object is being referred to as cat instead of self, the existing version didn't actually change the _codes of the resulting object. The segfault would occur when try to view the resulting Categorical, as you'd have take_1d trying to take out of bounds _codes here:

pandas/pandas/core/arrays/categorical.py

Line 1297 in caf462c

ret = take_1d(self.categories.values, self._codes)

[jschendel]

This change shouldn't alter existing behavior, as self and cat should be identical within this branch of the if/else. More so a defensive future-proofing change in case future modifications cause a divergence between self and cat prior to this. All operations at this point should be in relation to cat anyways, as that's the object we'll be returning.

[jreback]

thanks @jschendel