[vLLM] Parametrize embedding benchmarks into EMBEDDING_CONFIGS

[jazpurTT]

Problem Description

• Embedding benchmarks were defined as one hand-written test function per model/batch variant (test_vllm_bge_m3_batch1, test_vllm_bge_m3_batch32, test_vllm_qwen3_embedding_4b_batch1), unlike generative models which parametrize over the SINGLE_DEVICE_CONFIGS / TP_CONFIGS arrays. This doesn't scale as we expand the embedding config matrix — every new model/config would need another bespoke function.

What's Changes

• Added an EMBEDDING_CONFIGS array (mirroring SINGLE_DEVICE_CONFIGS / TP_CONFIGS) with stable pytest.param(..., id=...) IDs.
• Replaced the three per-model functions with a single parametrized test_vllm_embedding_benchmark.
• Updated the three embedding pytest node IDs in .github/workflows/perf-bench-matrix.json to the new parametrized form (e.g. test_vllm_embedding_benchmark[bge-m3-batch1]).
• No behavior change: same 3 configs and args; pure refactor to ready the infra before adding new embedding models/configs.

Checklist

• New/Existing tests provide coverage for changes

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 33.82%. Comparing base (9b0c875) to head (7280820).
⚠️ Report is 6 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5473      +/-   ##
==========================================
- Coverage   33.84%   33.82%   -0.03%     
==========================================
  Files          37       37              
  Lines        4990     4990              
==========================================
- Hits         1689     1688       -1     
- Misses       3301     3302       +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.