[CK2/CM scripts] support for ML frameworks and main ToDos

[x] Add readme.md in a script to describe limitations (such as in TF)
[x] Check that we delete local env inside dependencies loop
[x] Add -enable_if_env flag and --skip_if_env in script automation similar to CK
[x] Detect TF via Python first; check version ranges; and install only if missing ...
[x] Check virtual env creation based on Python dependency as a script (can env be automated?)
[x] Check caching of deps: https://github.com/mlcommons/ck/issues/248
[x] Fix contamination of ENV when detecting/installing artifacts
[x] ~~Should we do get/install as variation?~~ Possible but we keep it for later if needed by the community
[x] Check that can add extra tags from Python (including virtual + user tags) to other scripts with Python dependency
[x] Check detect/install of ONNX-RT, PyTorch, TVM, Transformers
[ ] Add basic MLPerf image classification benchmark workflow (onnv/pytorch/tf/tvm)
- [ ] Add Docker container for this implementation
[x] Add "cm show cache" (print tags, new ENV, versions from ENV)
[x] Add dummy "cm test script"
[x] Add "cm info" to get the {automation UID}::{artifact UID} for a given CM artifact
[x] Test "get-gcc" with corner detection script
- [x] See issue when both GCC and LLVM are detected: https://github.com/mlcommons/ck/issues/246
- [x] Separate compile and run scripts
[ ] Add support for Android devices and test corner detection
[ ] Check platform.init scripts from CK (Android and other devices)
[ ] Add script for CUDA detection
- [ ] Check that can work with different ranges of GCC versions on Linux
- [x] Create a simple script to detect CUDA capabilities (compile)
[x] Rename mcommons@cm-devops to cm-mlops ?
[ ] Agree how to create "cm test script"
[x] Move stable scripts to mlcommons@cm-mlops when ready ...
[ ] ~~Update docs, architecture and basic tutorial~~: Moved to https://github.com/mlcommons/ck/issues/256

mlcommons / ck

[CK2/CM scripts] support for ML frameworks and main ToDos #245