Report GPU power limit using the correct NVML API

[amarathe84]

Description

This PR replaces the nvmlDeviceGetPowerManagementLimit() call by the nvmlDeviceGetEnforcedPowerLimit(). The former API does not report the updated GPU power limit if it has been modified via. an out-of-band interface. For example, if the GPU power limit is implicitly updated by the node power limiting interface on Lassen, the updated GPU power limit is not reflected in the output of the former API. The latter API (i.e., nvmlDeviceGetEnforcedPowerLimit()) captures the changes in the GPU power limit applied through both the in-band interface (e.g., nvidia-smi or NVML) as well as the out-of-band interface (e.g., IBM node power limit interface).

Fixes #445

Type of change

• Bug fix (non-breaking change which fixes an issue)

How Has This Been Tested?

This fix has been tested on Lassen. Following test cases have been performed:

• Test 0: Report default GPU power limit. Result: Pass
• Test 1: Report GPU power limit after changing it with nvidia-smi (i.e., in-band). Result: Pass
• Test 2: Report GPU power limit after resetting it to default (300 W) with nvidia-smi (in-band). Result: Pass
• Test 3: Report GPU power limit after implicitly changing it with the IBM node power limit interface (out-of-band). Result: Pass
• Test 4: Report GPU power limit after implicitly resetting it with IBM node power limit interface (out-of-band). Result: Pass

Checklist:

• I have run ./scripts/check-code-format.sh and confirm my code code follows the style guidelines of variorum
• My changes generate no new warnings (build with -DENABLE_WARNINGS=ON)
• New and existing unit tests pass with my changes
• Added failure checks previously missing for the API call

Thank you for taking the time to contribute to Variorum!