Refactor Dockerfile-cuda-all into compute-cap stages

[alvarobartt]

What does this PR do?

This PR refactors the Dockerfile-cuda-all image to build one router binary per supported compute capability in a dedicated build stage, instead of building all variants from a shared mutable CUDA toolchain state.

Whilst moving from dynamic-linking to static-linking + nvprune fixes the startup delay of ~3 minutes when running the ...-cuda-1.9 container, it also implies that the image is larger given that it needs to prune and statically link CUDA to every binary.

Fixes #837

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline?
• Was this discussed/approved via a GitHub issue or the forum? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the documentation guidelines.
• Did you write any new necessary tests? If applicable, did you include or update the insta snapshots?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR.