[BUG] GroupBy Transform return incorrect results

[cdeotte]

Describe the bug
cuDF's groupby transform returns incorrect results

Steps/Code to reproduce bug
I have noticed that when the index contains non-consecutive numbers, then cuDF groupby transform returns incorrect results.

Expected behavior
cuDF groupby transform should return correct results

Environment overview (please complete the following information)
Version cuDF is 22.04.00

Additional context
Here is a jupyter notebook to reproduce the error
https://github.com/cdeotte/RAPIDS-development/blob/master/bugs/bug005.ipynb

[shwina]

The issue may simply be that we discard that index during a groupby transform, while Pandas maintains it:

In [13]: df
Out[13]:
   a  b
7  2  1
6  4  2
5  1  3
4  1  4
3  2  5
2  3  6
1  1  7

In [14]: df.groupby('a').transform('max')
Out[14]:
   b
0  5
1  2
2  7
3  7
4  5
5  6
6  7

In [15]: df.to_pandas().groupby('a').transform('max')
Out[15]:
   b
7  5
6  2
5  7
4  7
3  5
2  6
1  7

[bdice]

@shwina I think we may be susceptible to this class of error (dropping an input’s index or name when creating the output) in other places too. I caught a similar problem in #10715. If you can determine whether this is (1) not being tested or (2) not being caught by tests, that would be helpful.

(edit: this problem is slightly different than the class of error I had in mind -- nothing special is needed here. We're probably fine.)

[shwina]

I verified that #11068 fixes this issue.