Performance regression in series_methods.Clip.time_clip

[TomAugspurger]

https://pandas.pydata.org/speed/pandas/index.html#series_methods.Clip.time_clip?python=3.8&Cython=0.29.21&p-n=50&commits=29512dc9ea2b066ab5f4a7cb80c1f8f5becef1c2

#41141

I don't know if there's any easy way to get this back, but it'd be nice to avoid the slowdown if possible.

cc @Rasori

[TLouf]

I think the .to_numpy conversion can safely be removed to improve performance, I've benchmarked it and got

       before           after         ratio
     [b2a36bdf]       [c02f0362]
     <master>         <perf-time-clip>
-        605±50μs         483±10μs     0.80  series_methods.Clip.time_clip(1000)

with all tests in "frame/methods/test_clip.py" and "series/methods/test_clip.py" still passing. If this sounds ok I can write a PR.

[simonjayhawkins]

@TLouf can you open a PR with the proposed changes

[TLouf]

Done

[simonjayhawkins]

Thanks @TLouf