Decouple pandera and pandas when using it with polars

[glofru]

Is your feature request related to a problem? Please describe.

I am using pandera for a project with polars. I noticed that, installing pandera, pandas is installed too, since it is a requirement for pandera. The impact is an increased build size of 80+MB (size of pandas, numpy and pyarrow packages) that polars do not need, as well as pandera when checking polars DataFrame. I tried to uninstall pandas manually with pip and, despite all the imports in my package are either pandera.polars or pandera.typing.polars, the runtime fails because pandas is not installed. This because, in the pandera __init__ file, it imports pandera.backends, that eventually imports pandas.

Describe the solution you'd like
I would like to use pandera with polars without installing pandas.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.

[baldwinj30]

One thing that might be interesting along these lines is the narwhals package.

I have not taken too much of a deep dive into it, but from what I've seen using their API could help build a unified back-end that doesn't require any one framework.