fix: Resolve clippy warnings in source and test files

- Replace or_insert_with(Vec::new) with or_default() (src/auto_discovery.rs, src/discovery.rs)
- Prefix unused variables with underscore (tests)
- Remove redundant serde_json import (tests/vision_tests.rs)

Author: Michael-A-Kuykendall

src/auto_discovery.rs

@@ -305,7 +305,7 @@ impl ModelAutoDiscovery {
                     let group_key = format!("{}{}", base_name, extension);
                     shard_groups
                         .entry(group_key)
-                        .or_insert_with(Vec::new)
+                        .or_default()
                         .push(file_path.clone());
                     processed_files.insert(file_path.clone());
                 }

src/discovery.rs

@@ -183,7 +183,7 @@ impl ModelDiscovery {
                     );
                     shard_groups
                         .entry(group_key)
-                        .or_insert_with(Vec::new)
+                        .or_default()
                         .push(file_path.clone());
                     processed_files.insert(file_path.clone());
                 } else {

tests/regression/issue_142_amd_gpu_detection.rs

@@ -61,7 +61,7 @@ mod issue_142_tests {
 
         // Just verify that CPU backend creation doesn't panic and works correctly
         let _engine = shimmy::engine::llama::LlamaEngine::new_with_backend(Some("cpu"));
-        assert!(true); // If we get here, the test passes
+        // Test passes if we reach here without panicking
     }
 
     #[test]
@@ -80,9 +80,9 @@ mod issue_142_tests {
         let _engine = shimmy::engine::llama::LlamaEngine::new_with_backend(Some("auto"));
 
         // At least one GPU variable should be set if GPU backends are available
-        let has_cuda = env::var("GGML_CUDA").is_ok();
-        let has_vulkan = env::var("GGML_VULKAN").is_ok();
-        let has_opencl = env::var("GGML_OPENCL").is_ok();
+        let _has_cuda = env::var("GGML_CUDA").is_ok();
+        let _has_vulkan = env::var("GGML_VULKAN").is_ok();
+        let _has_opencl = env::var("GGML_OPENCL").is_ok();
 
         // If any GPU backend is enabled, at least one variable should be set
         #[cfg(any(

tests/vision_api_integration.rs

@@ -569,6 +569,6 @@ mod vision_disabled_tests {
     fn test_vision_feature_disabled() {
         // When vision feature is disabled, these tests should not run
         println!("Vision feature disabled - integration tests skipped");
-        assert!(true);
+        // Test passes if we reach here without panicking
     }
 }

tests/vision_tests.rs

@@ -14,7 +14,6 @@ mod vision_tests {
     use super::*;
     use base64::{engine::general_purpose, Engine as _};
     use image::{codecs::png::PngEncoder, ColorType, ImageEncoder};
-    use serde_json;
     use serial_test::serial;
     use std::collections::HashMap;