BUG/TST: can_use_libjoin cases with mixed-freq Periods

[jbrockmendel]

In the Index code we have a fastpaths in for _union, _intersection and join, that look something like

if self._can_use_libjoin and [...]
     try:
        result = self._use_fastpath(...)
     except TypeError:
         # object dtype, non-comparable objects
         result = self._non_fastpath(...)

_union also catches IncompatibleFrequency (which subclasses ValueError instead of TypeError, which itself might be something to somehow deprecate) which is what you get if you try to sort Period objects with mismatched freqs.

I'm guessing that the _intersection and join paths should also catch IncompatibleFrequency. Either that or check that we have test cases that get there with mismatches Periods inside object dtype indexes.

[jbrockmendel]

ts = pd.Timestamp.now()
pers = [ts.to_period(freq) for freq in ["D", "s", "h", "M"]]

left = pd.Index(pers[:2])
right = pd.Index(pers[2:])

>>> left.intersection(right)
Traceback (most recent call last):
[...]
pandas._libs.tslibs.period.IncompatibleFrequency: Input has different freq=D from Period(freq=s)

>>> left.join(right)
Traceback (most recent call last):
[...]
pandas._libs.tslibs.period.IncompatibleFrequency: Input has different freq=D from Period(freq=s)

[AmeyaVerma]

I can fix this please assign me with it