Auto-generate distro yamls + docs (#468)

# What does this PR do?

Automatically generates
- build.yaml
- run.yaml
- run-with-safety.yaml
- parts of markdown docs

for the distributions.

## Test Plan

At this point, this only updates the YAMLs and the docs. Some testing
(especially with ollama and vllm) has been performed but needs to be
much more tested.

llama_stack/providers/inline/inference/meta_reference/config.py

@@ -49,6 +49,18 @@ class MetaReferenceInferenceConfig(BaseModel):
         resolved = resolve_model(self.model)
         return resolved.pth_file_count
 
+    @classmethod
+    def sample_run_config(
+        cls,
+        model: str = "Llama3.2-3B-Instruct",
+        checkpoint_dir: str = "${env.CHECKPOINT_DIR:null}",
+    ) -> Dict[str, Any]:
+        return {
+            "model": model,
+            "max_seq_len": 4096,
+            "checkpoint_dir": checkpoint_dir,
+        }
+
 
 class MetaReferenceQuantizedInferenceConfig(MetaReferenceInferenceConfig):
     quantization: QuantizationConfig

llama_stack/providers/inline/inference/meta_reference/generation.py

@@ -107,7 +107,7 @@ class Llama:
             sys.stdout = open(os.devnull, "w")
 
         start_time = time.time()
-        if config.checkpoint_dir:
+        if config.checkpoint_dir and config.checkpoint_dir != "null":
             ckpt_dir = config.checkpoint_dir
         else:
             ckpt_dir = model_checkpoint_dir(model)
@@ -137,7 +137,6 @@ class Llama:
         ), f"model_args vocab = {model_args.vocab_size} but tokenizer vocab = {tokenizer.n_words}"
 
         if isinstance(config, MetaReferenceQuantizedInferenceConfig):
-
             if isinstance(config.quantization, Fp8QuantizationConfig):
                 from .quantization.loader import convert_to_fp8_quantized_model