[SPARK-40577][PS] Fix CategoricalIndex.append to match pandas 1.5.0

[itholic]

What changes were proposed in this pull request?

The PR proposes to fix CategoricalIndex.append to match the behavior with pandas.

Why are the changes needed?

Because the current behavior is different from pandas 1.5.0.

Does this PR introduce any user-facing change?

The behavior of API is changed as below:

Before

>>> psidx1 = ps.CategoricalIndex(["x", "y", "z"], categories=["z", "y", "x", "w"])
>>> psidx3 = ps.Index(["y", "x", "w", "z"])
>>> psidx1.append(psidx3.astype("category"))
CategoricalIndex(['x', 'y', 'z', 'y', 'x', 'w', 'z'], categories=['z', 'y', 'x', 'w'], ordered=False, dtype='category')

After

>>> psidx1 = ps.CategoricalIndex(["x", "y", "z"], categories=["z", "y", "x", "w"])
>>> psidx3 = ps.Index(["y", "x", "w", "z"])
>>> psidx1.append(psidx3.astype("category"))
CategoricalIndex(['x', 'y', 'z', 'x', 'y', 'z', 'w'], categories=['z', 'y', 'x', 'w'], ordered=False, dtype='category')

How was this patch tested?

Manually check the existing test is passed with pandas 1.5.0.

[HyukjinKwon]

Merged to master.

[zhengruifeng]

Merged to master.

@HyukjinKwon this PR seems not merged?

[HyukjinKwon]

oops