feat: launcher auto-adds --host/--port; recipes no longer hardcode

The launcher now auto-adds '--host 0.0.0.0' and '--port <recipe.port>'
for both vLLM and SGLang.  Recipes no longer need to specify these.

- Removed '--host 0.0.0.0' and '--port 30000' from the DeepSeek V4
  recipe args (now provided by the launcher)
- Port is driven by `recipe.port` (default 30000).  Recipes can
  override via the `port` field if they need a different value.

recipes/deepseek-v4-flash/recipe.toml

@@ -2,7 +2,6 @@ engine = "vllm"
 name = "DeepSeek V4 Flash (vLLM)"
 docker_image = "aidendle94/sparkrun-vllm-ds4-gb10:production-ready"
 model = "deepseek-ai/DeepSeek-V4-Flash"
-port = 8000
 required_nodes = 2
 tensor_parallel_size = 2
 pipeline_parallel_size = 1
@@ -29,8 +28,6 @@ args = [
     "--default-chat-template-kwargs.thinking", "true",
     "--default-chat-template-kwargs.reasoning_effort", "high",
     "--enable-flashinfer-autotune",
-    "--host", "0.0.0.0",
-    "--port", "8000",
     "--trust-remote-code",
     "--served-model-name", "deepseek-v4-flash",
 ]

src/docker/mod.rs

@@ -471,11 +471,12 @@ fn build_docker_command(
     }
     cmd.extend(flags);
 
-    // Port flag — auto-generated for SGLang (all nodes need to bind)
-    if is_sglang {
-        cmd.push("--port".to_string());
-        cmd.push(recipe.port.to_string());
-    }
+    // Host + port flags — auto-generated for both engines so recipes
+    // don't need to hardcode them.
+    cmd.push("--host".to_string());
+    cmd.push("0.0.0.0".to_string());
+    cmd.push("--port".to_string());
+    cmd.push(recipe.port.to_string());
 
     cmd
 }