Add AMDSMIObserver that uses amdsmi to measure energy

[stijnh]

This PR adds AMDSMIObserver, a new observer that uses amdsmi to monitor the GPU.

It can measure:

• energy (in Joule)
• core clock frequency
• memory clock frequency
• temperature
• core voltage

It requires no dependencies, besides amdsmi (which is installed by default with rocm).

I have tested this on LUMI with ROCm 6.3.4. The ROCm API is quite unstable so hopefully it also works with other versions.

[sonarqubecloud]

Quality Gate passed

Issues
10 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[wjp]

It seems to work with rocm 7.1 for me as well.

I do get an error if I don't pass device_id=0 to the AMDSMIObserver (ValueError: failed to detect AMD device: invalid UUID of backend: None), when testing with examples/hip/vector_add.py . I don't see HipFunctions storing the uuid, so maybe I'm not using the right version of the KT hip backend yet?