cuda: Verify counter and throughput metrics are supported on each unique chip

[Treece-Burgess]

Pull Request Description

This PR filters the Perfworks counter and throughput metrics by verifying that they are supported on the unique chip and by doing so resolves Issue #542.

It is important to note that this change results in longer execution time.

Testing

Testing was done on Heimdall (1 * RTX5080) at Oregon with Cuda Toolkit 12.9 and Guyot (8 * A100) at ICL with Cuda Toolkit 12.9. The configure used is as follows:

./configure --prefix=$PWD/test-install --with-components="cuda" --with-debug=yes

Heimdall

• PAPI build: ✅
• PAPI utilities*: ✅
• Cuda component tests: ✅
• Issue 542: ✅
• Wall clock time comparison of papi_component_avail
  • This PR: [2.464, 2.492, 2.488]
  • Master branch: [0.463, 0.398, 0.429]

Guyot

• PAPI build: ✅
• PAPI utilities*: ✅
• Cuda component tests: ✅
• Wall clock time comparison of papi_component_avail
  • This PR: [8.058, 8.011, 8.068]
  • Master branch: [5.004, 5.024, 4.998]

* - papi_component_avail, papi_native_avail, papi_command_line

Author Checklist

• Description
  Why this PR exists. Reference all relevant information, including background, issues, test failures, etc
• Commits
  Commits are self contained and only do one thing
  Commits have a header of the form: module: short description
  Commits have a body (whenever relevant) containing a detailed description of the addressed problem and its solution
• Tests
  The PR needs to pass all the tests