Potential performance issue: Unreliable performance of df.ewm in pandas 1.2.2

[TendouArisu]

Issue Description:

Hello.
I have discovered a performance degradation in the df.ewm function of pandas version 1.2.2 when ewm passed with times and it is so slow compared to ewm pased with int. And I notice that requirements.txt, shows that it depends on pandas version 1.2.2. I am not sure whether this performance problem in pandas will affect this repository. I found some discussions on GitHub related to this issue, including #39784 and #40072.
I also found that 06_rnns.ipynb used the influenced api. There may be more files used the influenced api.

Suggestion

I would recommend considering an upgrade to a different version of pandas > 1.2.2 or exploring other solutions to optimize the performance.
Any other workarounds or solutions would be greatly appreciated.
Thank you!