[CI][GPU][C++][Python] Add new jobs for cuda

[raulcd]

Describe the enhancement requested

We used to have cuda runners provided by voltron data infrastructure. We have lost access to those and have started working on a new solution.
Using the project AWS account we have configured with ASF infra a runs-on solution.

We tested on a draft PR that the runners were available:
#48415

This issue is to clean the old jobs that are not available anymore and add new runners.

Component(s)

C++, Continuous Integration, GPU, Python

[raulcd]

Issue resolved by pull request 48583
#48583