BUG: GH10355 groupby std() no longer sqrts grouping cols

[henrystokeley]

closes #10355

Previously grouping columns were square rooted when as_index=False
New method closely follows the format of var() method.

[jreback]

see comments on #11085 w.r.t. performance of this

[henrystokeley]

My solution is slightly different from #11085 (I keep the distinction from ddof == 1 and ddof != 1)
I believe my solution is essentially the same as the sketch one proposed on #10355 where you calculate the solution and then take the square root.

Isn't this the same performance-wise as the current method which just square roots the answer from var() (but also includes unnecessary columns)
I'm not an expert on performance so could you clarify this for me?

[jreback]

the problem is that when you go to a python soln it is evaluated for every group rather than using the cython function. The soln needs to simply take the .var (with the correct ddof), then np.sqrt on the result (excluding the selection columns).

[jreback]

can you update according to comments

[jreback]

@henrystokeley can you update according to comments

[jreback]

replaced by #11507

[xieyuheng]

#25315