Add Qwen3-Coder-30B setup documentation to activity37

Added detailed comments showing how to use the recommended model:
- hf.co/unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF:Q4_K_M
- Setup instructions for llama.cpp (with GPU offloading)
- Alternative setup with ollama
- Environment variable configuration examples

This 30B parameter model is specifically optimized for code generation
across all programming languages, making it perfect for the universal
programming activity.

research/activity37-programming-languages.yaml

@@ -1,8 +1,24 @@
 default_max_attempts_per_step: 3
 
 # Model configuration
-# Use Hermes for classification (fast, accurate bucketing)
-# Use Qwen 3 Coder for feedback (specialized for code generation)
+# classifier_model: Fast classification into buckets (correct, partial, etc.)
+#   MODEL_1 = Hermes-3-Llama-3.1-8B (always available, great for role-play)
+#
+# feedback_model: Code generation and feedback
+#   MODEL_3 = Qwen3-Coder-30B-A3B-Instruct (specialized for code)
+#   Recommended: hf.co/unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF:Q4_K_M
+#
+#   Setup with llama.cpp:
+#     1. Download: huggingface-cli download unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF \
+#                    Qwen3-Coder-30B-A3B-Instruct-Q4_K_M.gguf
+#     2. Run: llama-server -m Qwen3-Coder-30B-A3B-Instruct-Q4_K_M.gguf \
+#               --host 0.0.0.0 --port 8080 -ngl 99
+#     3. Set env: export MODEL_ENDPOINT_3=http://localhost:8080/v1
+#                 export MODEL_API_KEY_3=dummy
+#
+#   Or use with ollama:
+#     ollama run unsloth/qwen3-coder:30b-instruct-q4_K_M
+#
 classifier_model: "MODEL_1"
 feedback_model: "MODEL_3"