GPU Standalone benchmark can compile with ONNX + couple of related fixes

[davidrohr]

@ChSonnabend : I checked a bit the ONNXCode. This PR fixes a couple of compiler warnings due to signed vs unsigned comparisons, shadowed variables, unchecked function call results.

Also, I think the CUDA ORT code did never work, and it is actually not compiled in the CI, since we have only the ONNXRuntime for ROCm there.
I fixed the problem that the code used the calls of the OrtApi struct directly, without getting the OrtApi first.
This is fixed now, but now it fails for me since OrtCUDAProviderOptionsV2 is not defined, and I couldn't easily find out what is wrong.

Could you check the ORT CUDA code?
Also, I saw a lot of int / uint in the NN Clusterizer code, while the GPU code should only use int32_t / uint32_t. Could you change that?
Also, the member variables names of the NN Clusterizer do not follow the naming convention: https://rawgit.com/AliceO2Group/CodingGuidelines/master/naming_formatting.html

• normal member variables must start with m
• static variables with s
• CONSTEXPR variables must be capitalized and have no prefix (IIRC).
  Could you adapt that as well?

[github-actions]

REQUEST FOR PRODUCTION RELEASES:
To request your PR to be included in production software, please add the corresponding labels called "async-" to your PR. Add the labels directly (if you have the permissions) or add a comment of the form (note that labels are separated by a ",")

+async-label <label1>, <label2>, !<label3> ...

This will add <label1> and <label2> and removes <label3>.

The following labels are available
async-2023-pbpb-apass4
async-2023-pp-apass4
async-2024-pp-apass1
async-2022-pp-apass7
async-2024-pp-cpass0
async-2024-PbPb-apass1
async-2024-ppRef-apass1
async-2024-PbPb-apass2
async-2023-PbPb-apass5

[ChSonnabend]

Sure, I'll have a look. Thanks a lot for all the changes!