Address local model download review

Return immediately from the server download endpoint and preserve terminal cancellation state for HF-backed downloads. Mark pre-download failures in the shared progress map so clients can surface actionable errors.

crates/goose/src/providers/local_inference/hf_models.rs

@@ -1742,7 +1742,7 @@ impl ProgressHandler for HfDownloadProgress {
                     update_download_manager_progress(&self.model_id, &state);
                 }
             }
-            DownloadEvent::Complete => self.complete(),
+            DownloadEvent::Complete => {}
         }
     }
 }

crates/goose/src/providers/local_inference/local_model_registry.rs

@@ -206,20 +206,15 @@ pub fn get_registry() -> &'static Mutex<LocalModelRegistry> {
     })
 }
 
-#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize, ToSchema)]
+#[derive(Debug, Clone, Copy, Default, PartialEq, Eq, Serialize, Deserialize, ToSchema)]
 #[serde(rename_all = "snake_case")]
 pub enum LocalModelStorage {
+    #[default]
     GooseManaged,
     HuggingFaceCache,
     ManualPath,
 }
 
-impl Default for LocalModelStorage {
-    fn default() -> Self {
-        Self::GooseManaged
-    }
-}
-
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct ShardFile {
     pub filename: String,

crates/goose/src/providers/local_inference/mlx.rs

@@ -367,8 +367,6 @@ mod imp {
 
 #[cfg(not(feature = "mlx"))]
 mod imp {
-    use std::any::Any;
-
     use crate::providers::errors::ProviderError;
     use crate::providers::local_inference::backend::{
         BackendLoadedModel, LocalGenerationRequest, LocalInferenceBackend,
@@ -386,14 +384,6 @@ mod imp {
         }
     }
 
-    struct UnavailableLoadedModel;
-
-    impl BackendLoadedModel for UnavailableLoadedModel {
-        fn as_any_mut(&mut self) -> &mut dyn Any {
-            self
-        }
-    }
-
     impl LocalInferenceBackend for MlxBackend {
         fn id(&self) -> &'static str {
             MLX_BACKEND_ID