[SPARK-55296][PS][FOLLOW-UP] Fix CoW mode not to break groupby

[ueshin]

What changes were proposed in this pull request?

This is a follow-up of #54375.

Fixes CoW mode not to break groupby.
Delays to disconnect the anchor to when actually being updated.

Why are the changes needed?

The CoW mode was supported at #54375, but it disconnected the anchor too early, causing to break groupby.

>>> import pandas as pd
>>> import pyspark.pandas as ps
>>>
>>> pdf1 = pd.DataFrame({"C": [0.362, 0.227, 1.267, -0.562], "B": [1, 2, 3, 4]})
>>> pdf2 = pd.DataFrame({"A": [1, 1, 2, 2]})
>>>
>>> psdf1 = ps.from_pandas(pdf1)
>>> psdf2 = ps.from_pandas(pdf2)
>>>
>>> pdf1.groupby([pdf1.C, pdf2.A]).agg("sum").sort_index()
          B
C      A
-0.562 2  4
 0.227 1  2
 0.362 1  1
 1.267 2  3
>>> psdf1.groupby([psdf1.C, psdf2.A]).agg("sum").sort_index()
              C  B
C      A
-0.562 2 -0.562  4
 0.227 1  0.227  2
 0.362 1  0.362  1
 1.267 2  1.267  3

Does this PR introduce any user-facing change?

Yes, it will behave more like pandas 3.

How was this patch tested?

The existing tests should pass.

Was this patch authored or co-authored using generative AI tooling?

Codex (GPT-5.3-Codex)

[ueshin]

cc @gaogaotiantian @HyukjinKwon @zhengruifeng

[HyukjinKwon]

Merged to master.