feat(test): RunPod GPU cost attribution demo + test Dockerfile

Dockerfile.test: builds agent from source + PyTorch cu124 — no PyPI
release needed; works on any RunPod GPU Pod.

agent/tests/runpod_demo.py: end-to-end demo that proves the cost
attribution story:
  Phase 1 — idle baseline: sample GPU power with no workload;
    show projected monthly cost of leaving GPUs idle unmonitored
  Phase 2 — fp32 dense matmul: saturate CUDA cores, record TFLOPS
  Phase 3 — bf16 matmul: tensor core path, compare power vs TFLOPS
  Phase 4 — token generation: simulated attention loop, tokens/s
  Summary table: avg W, baseline-adjusted W, kWh/TFLOP, $/hr, $/month
  --upload flag: submits best result to Green AI Index leaderboard

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: AgentMulder404

Dockerfile.test

@@ -0,0 +1,54 @@
+# AluminatiAI — RunPod test image
+#
+# Builds the agent from source (no PyPI release needed) and bundles
+# PyTorch so the demo workload can actually saturate the GPU.
+#
+# Build:
+#   docker build -f Dockerfile.test -t aluminatai-test .
+#
+# Run on RunPod GPU Pod (pick any CUDA-capable template, then SSH in and run):
+#   docker run --rm --gpus all --pid=host \
+#     -e ALUMINATAI_API_KEY=alum_... \
+#     -e ALUMINATAI_TEAM=demo \
+#     -e DRY_RUN=0 \
+#     -e METRICS_PORT=9100 \
+#     -p 9100:9100 \
+#     aluminatai-test python3 /app/tests/runpod_demo.py
+#
+# Or to just drop into a shell and poke around:
+#   docker run --rm -it --gpus all --pid=host aluminatai-test bash
+
+FROM nvidia/cuda:12.4.1-runtime-ubuntu22.04
+
+ENV DEBIAN_FRONTEND=noninteractive \
+    PYTHONUNBUFFERED=1 \
+    PYTHONDONTWRITEBYTECODE=1
+
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends \
+        python3-pip \
+        curl \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+# Install PyTorch (CPU+CUDA wheel — lets us run GPU workloads for the demo)
+RUN pip3 install --no-cache-dir \
+    torch --index-url https://download.pytorch.org/whl/cu124
+
+# Install agent dependencies + prometheus support
+COPY agent/pyproject.toml /app/
+COPY agent/ /app/
+RUN pip3 install --no-cache-dir \
+    "nvidia-ml-py>=12.0.0" \
+    "requests>=2.28" \
+    "python-dotenv>=1.0" \
+    "rich>=13.0" \
+    "prometheus-client>=0.19" \
+    "numpy>=1.24"
+
+# Copy test scripts
+COPY agent/tests/ /app/tests/
+
+# Default: run the demo
+CMD ["python3", "/app/tests/runpod_demo.py"]