feat: cuDF backend

[lostmygithubaccount]

Is your feature request related to a problem?

while we verified (at least on one version of Ibis) that cuDF "just works" with the pandas backend, it'd be good to have a proper dedicated and tested cuDF backend for a couple reasons:

1. a better story for IbisML
2. a better story for Theseus

"Use Ibis and IbisML to accelerate your ML preprocessing and XGBoost/LightGBM/PyTorch/Tensorflow training on a single GPU node. If hitting a wall, scale up w/ Theseus"

Describe the solution you'd like

proper cuDF backend w/ testing. probably need a GPU

documentation

blog

What version of ibis are you running?

main

What backend(s) are you using, if any?

n/a

Code of Conduct

• I agree to follow this project's Code of Conduct

[kszucs]

With the new pandas executor we are in a better position to optimize for cuDF now. What are the options for a CI setup?

[gforsyth]

cc @mmccarty -- y'all have any pointers on cuDF CI setups?

[gforsyth]

I'm going to close this out for now -- we might pick it back up in the future but it isn't a current priority.