feat(ci): test lookup on cpu without neuronx

This is to veirfy that, disregarding what it is done with commands, it
is always possible to install optimum neuron without the neuronx extra
and run the cache lookup command.

Author: tengomucho

.github/workflows/test_cpu_lookup.yml

@@ -0,0 +1,60 @@
+name: Optimum Neuron / Cache Lookup on CPU
+
+on:
+  push:
+    branches: [ main ]
+    paths:
+      - "pyproject.toml"
+      - "optimum/**.py"
+      - ".github/workflows/test_cpu_lookup.yml"
+  pull_request:
+    branches: [ main ]
+    paths:
+      - "pyproject.toml"
+      - "optimum/**.py"
+      - ".github/workflows/test_cpu_lookup.yml"
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
+
+jobs:
+  test-cpu-compilation:
+    name: Run CPU Compilation Tests
+    runs-on: ubuntu-22.04
+    env:
+      MODEL_ID: Qwen/Qwen3-0.6B
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Setup Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: "3.11"
+      - name: Install Neuronx runtime
+        uses: ./.github/actions/install_neuronx_runtime
+      - name: Prepare virtual environment
+        uses: ./.github/actions/prepare_venv
+      - name: Install optimum-neuron
+        uses: ./.github/actions/install_optimum_neuron
+      - name: Setup PATH
+        run: echo "/home/ubuntu/.local/bin" >> $GITHUB_PATH
+      - name: Export sample model and synchronize cache
+        run: |
+          source aws_neuron_venv_pytorch/bin/activate
+          HF_TOKEN=${{ secrets.HF_TOKEN_OPTIMUM_NEURON_CI }} \
+            optimum-cli export neuron -m $MODEL_ID /tmp/exported-$MODEL_ID \
+            --batch_size 4 --sequence_length 4096 --tensor_parallel_size 2 --instance_type trn1
+          HF_TOKEN=${{secrets.HF_TOKEN_OPTIMUM_NEURON_CACHE}} \
+            optimum-cli neuron cache synchronize
+      - name: Setup new virtual environment without neuronx
+        run: |
+          python3 -m venv on-no-neuronx
+          source on-no-neuronx/bin/activate
+          pip install torch==2.8.0 torchvision~=0.23 --index-url https://download.pytorch.org/whl/cpu
+          pip install .
+      - name: Run lookup test
+        run: |
+          source on-no-neuronx/bin/activate
+          HF_TOKEN=${{secrets.HF_TOKEN_OPTIMUM_NEURON_CACHE}} \
+            optimum-cli neuron cache lookup $MODEL_ID