Clean up groupby to use the anchor.

[ueshin]

Now that Series has a different anchor from the one it is originally derived from if it is modified, we don't need to check Spark column's equality in groupby.

Also fixing a bug when it needs compute.ops_on_diff_frames but the first group key is from the same DataFrame but modified.

E.g.,

>>> ks.options.compute.ops_on_diff_frames = True
>>> kdf1 = ks.DataFrame({"C": [0.362, 0.227, 1.267, -0.562], "B": [1, 2, 3, 4]})
>>> kdf2 = ks.DataFrame({"A": [1, 1, 2, 2]})
>>> kdf1.groupby([kdf1.C + 1, kdf2.A]).sum()
Traceback (most recent call last):
...
AttributeError: 'NoneType' object has no attribute '_internal'

[ueshin]

Let me merge this now. Please feel free to comment on this.

[HyukjinKwon]

Sorry for late review @ueshin. LGTM!