Add an awq_sym lm-eval test (#1702)

SUMMARY:
In response to
#1688, add an
awq_asym lm-eval test config.

TEST PLAN:
```
export CADENCE=weekly
export TEST_DATA_FILE=tests/lmeval/configs/w4a16_awq_asym.yaml
python -m pytest tests/lmeval/test_lmeval.py::TestLMEval::test_lm_eval -v
```

MISC:
As Llama is a gated model, I'm using a local model path instead of
meta-llama/Meta-Llama-3-8B-Instruct during local testing. I'm not sure
if this might affect the test flow / accuracy. It would be great if you
could help me verify it. That's also why I've marked it as WIP.

---------

Signed-off-by: aladerran <aladerran@gmail.com>
Co-authored-by: Dipika Sikka <dipikasikka1@gmail.com>

Author: aladerran

tests/e2e/vLLM/recipes/WNA16/recipe_w4a16_awq_sym.yaml

@@ -0,0 +1,13 @@
+quant_stage:
+  quant_modifiers:
+    AWQModifier:
+      ignore: ["lm_head"]
+      config_groups:
+        group_0:
+          weights:
+            num_bits: 4
+            type: "int"
+            symmetric: true
+            strategy: "group"
+            group_size: 128
+          targets: ["Linear"]

tests/lmeval/configs/w4a16_awq_sym.yaml

@@ -0,0 +1,10 @@
+cadence: "weekly"
+model: meta-llama/Meta-Llama-3-8B-Instruct
+scheme: W4A16
+recipe: tests/e2e/vLLM/recipes/AWQ/recipe_w4a16_awq_sym.yaml
+dataset_id: HuggingFaceH4/ultrachat_200k
+dataset_split: train_sft
+lmeval:
+  metrics:
+    exact_match,flexible-extract: 0.70
+    exact_match,strict-match: 0.70