run: mount all workload profiles in debug mode

[hubunt]

Summary

Fix run debug mode so the harness pod exposes workload profiles for all harnesses, not only the selected harness.

Motivation

Fixes #1272.

When llmdbenchmark run --debug is used, the harness pod is intended to stay alive for manual inspection and debugging. The current path only renders, creates, and mounts the
selected harness profile ConfigMap, so workloads from other harnesses are not available inside the debug pod.

What changed

• In debug mode, render profiles for every built-in harness under workload/profiles/*.
• In debug mode, create one profiles ConfigMap per rendered harness.
• In debug mode, mount every profiles ConfigMap under /workspace/profiles/<harness>.
• Use a generic debug pod name prefix: llmdbench-harness-debug-<suffix>.
• Preserve non-debug behavior: only the selected harness profile ConfigMap is rendered, created, and mounted.
• Added regression tests for debug profile rendering, ConfigMap creation, template mounts, and generic debug pod naming.

Testing

• python -m pytest tests/test_render_profiles_workload_file_path.py tests/test_create_profile_configmap.py tests/test_harness_pod_template_scripts.py tests/ test_deploy_harness_status.py -q
• python -m pytest tests -q
• python -m ruff format --check --diff ...
• python -m ruff check --unsafe-fixes --output-format=full ...
• python -m py_compile ...
• git diff --check

Full test result: 638 passed, 31 skipped.

Backward Compatibility

Non-debug run behavior is unchanged. Local --workload-file-path debug runs continue to render the explicitly provided workload file rather than expanding to all built-in
profiles.

[github-actions]

Unsigned commits detected! Please sign your commits.

For instructions on how to set up GPG/SSH signing and verify your commits, please see GitHub Documentation.