Fix compiler warnings in test files

Eliminate unused variable and import warnings across test files:
- streaming_parser_test.rs: prefix unused `tools` with underscore
- webdriver_session.rs: remove unused `use super::*` import
- mock_provider_integration_test.rs: prefix unused `result` and `task_result`
- test_preflight_max_tokens.rs: prefix unused `proposed_max`
- todo_staleness_test.rs: add #[allow(dead_code)] for test helper methods
- json_parsing_stress_test.rs: prefix unused `tools`
- read_file_token_limit_test.rs: add #[allow(dead_code)] for unused helper
- background_process_demo_test.rs: remove unused PathBuf import
- test_session_continuation.rs: prefix unused `temp_dir` in 7 tests

All tests pass. No behavior changes.

Agent: fowler

crates/g3-core/src/webdriver_session.rs

@@ -120,8 +120,6 @@ impl WebDriverSession {
 
 #[cfg(test)]
 mod tests {
-    use super::*;
-
     #[test]
     fn test_webdriver_session_enum_variants() {
         // This test just verifies the enum structure compiles correctly

crates/g3-core/tests/background_process_demo_test.rs

@@ -1,5 +1,4 @@
 use g3_core::background_process::BackgroundProcessManager;
-use std::path::PathBuf;
 use std::thread;
 use std::time::Duration;
 use std::fs;

crates/g3-core/tests/json_parsing_stress_test.rs

@@ -164,7 +164,7 @@ fn test_json_in_indented_code_block() {
 
 That's it."#;
     
-    let tools = parser.process_chunk(&chunk(content));
+    let _tools = parser.process_chunk(&chunk(content));
     parser.process_chunk(&finished_chunk());
     
     // Indented code blocks are trickier - the JSON is on its own line but indented

crates/g3-core/tests/mock_provider_integration_test.rs

@@ -331,7 +331,7 @@ async fn test_json_on_own_line_detected_as_tool() {
     // The task should detect the tool call
     // Note: This will fail because we don't have a real shell, but that's OK
     // We just want to verify the tool call was detected
-    let result = agent.execute_task("Run echo hello", None, false).await;
+    let _result = agent.execute_task("Run echo hello", None, false).await;
     
     // The result might be an error (tool execution fails in test env)
     // but we can check if a tool was attempted by looking at context
@@ -408,7 +408,7 @@ async fn test_response_not_duplicated() {
     assert!(result.is_ok(), "Task should succeed: {:?}", result.err());
 
     // Check the TaskResult - it should have the response
-    let task_result = result.unwrap();
+    let _task_result = result.unwrap();
     
     // The response field might be empty (content was streamed) or contain the response
     // Either way, the context should have exactly one assistant message with this content

crates/g3-core/tests/read_file_token_limit_test.rs

@@ -8,6 +8,7 @@ use std::path::PathBuf;
 use tempfile::TempDir;
 
 /// Helper to create a test file with specified size
+#[allow(dead_code)]
 fn create_test_file(dir: &TempDir, name: &str, size_bytes: usize) -> PathBuf {
     let path = dir.path().join(name);
     let content: String = "x".repeat(size_bytes);

crates/g3-core/tests/streaming_parser_test.rs

@@ -75,7 +75,7 @@ fn test_first_complete_second_incomplete() {
     let content = r#"{"tool": "read_file", "args": {"file_path": "a.txt"}}
 {"tool": "shell", "args": {"command": "ls"#;
     
-    let tools = parser.process_chunk(&chunk(content, false));
+    let _tools = parser.process_chunk(&chunk(content, false));
     
     // Should detect the first complete tool call
     // The incomplete one should be detected by has_incomplete_tool_call

crates/g3-core/tests/test_preflight_max_tokens.rs

@@ -34,7 +34,7 @@ fn test_no_thinking_budget_passes_through() {
     let config = create_test_config_with_thinking(None);
     
     // Without thinking budget, any max_tokens should be fine
-    let proposed_max = 5000;
+    let _proposed_max = 5000;
     
     // The constraint check would return (proposed_max, false)
     // since there's no thinking_budget_tokens configured

crates/g3-core/tests/test_session_continuation.rs

@@ -125,7 +125,7 @@ fn test_find_incomplete_agent_session() {
     use g3_core::session_continuation::find_incomplete_agent_session;
     
     let _lock = TEST_MUTEX.lock().unwrap();
-    let (temp_dir, original_dir) = setup_test_env();
+    let (_temp_dir, original_dir) = setup_test_env();
 
     // Get the actual current directory (after set_current_dir in setup)
     let current_working_dir = std::env::current_dir()
@@ -165,7 +165,7 @@ fn test_find_incomplete_agent_session_ignores_complete_todos() {
     use g3_core::session_continuation::find_incomplete_agent_session;
     
     let _lock = TEST_MUTEX.lock().unwrap();
-    let (temp_dir, original_dir) = setup_test_env();
+    let (_temp_dir, original_dir) = setup_test_env();
 
     let current_working_dir = std::env::current_dir()
         .map(|p| p.to_string_lossy().to_string())
@@ -197,7 +197,7 @@ fn test_find_incomplete_agent_session_ignores_non_agent_mode() {
     use g3_core::session_continuation::find_incomplete_agent_session;
     
     let _lock = TEST_MUTEX.lock().unwrap();
-    let (temp_dir, original_dir) = setup_test_env();
+    let (_temp_dir, original_dir) = setup_test_env();
 
     let current_working_dir = std::env::current_dir()
         .map(|p| p.to_string_lossy().to_string())
@@ -227,7 +227,7 @@ fn test_find_incomplete_agent_session_ignores_non_agent_mode() {
 #[test]
 fn test_load_continuation_when_none_exists() {
     let _lock = TEST_MUTEX.lock().unwrap();
-    let (temp_dir, original_dir) = setup_test_env();
+    let (_temp_dir, original_dir) = setup_test_env();
 
     // No continuation should exist in a fresh temp directory
     let result = load_continuation().expect("load_continuation should not error");
@@ -239,7 +239,7 @@ fn test_load_continuation_when_none_exists() {
 #[test]
 fn test_clear_continuation() {
     let _lock = TEST_MUTEX.lock().unwrap();
-    let (temp_dir, original_dir) = setup_test_env();
+    let (_temp_dir, original_dir) = setup_test_env();
 
     // Create and save a continuation
     let continuation = SessionContinuation::new(false, None, 
@@ -295,7 +295,7 @@ fn test_ensure_session_dir_creates_g3_directory() {
 #[test]
 fn test_has_valid_continuation_with_missing_session_log() {
     let _lock = TEST_MUTEX.lock().unwrap();
-    let (temp_dir, original_dir) = setup_test_env();
+    let (_temp_dir, original_dir) = setup_test_env();
 
     // Create a continuation pointing to a non-existent session log
     let continuation = SessionContinuation::new(false, None, 
@@ -347,7 +347,7 @@ fn test_has_valid_continuation_with_existing_session_log() {
 #[test]
 fn test_continuation_serialization_format() {
     let _lock = TEST_MUTEX.lock().unwrap();
-    let (temp_dir, original_dir) = setup_test_env();
+    let (_temp_dir, original_dir) = setup_test_env();
 
     let continuation = SessionContinuation::new(false, None, 
         "format_test".to_string(),

crates/g3-core/tests/todo_staleness_test.rs

@@ -22,14 +22,17 @@ impl MockUiWriter {
         }
     }
 
+    #[allow(dead_code)]
     fn set_prompt_response(&self, response: bool) {
         self.prompt_responses.lock().unwrap().push(response);
     }
 
+    #[allow(dead_code)]
     fn set_choice_response(&self, response: usize) {
         self.choice_responses.lock().unwrap().push(response);
     }
 
+    #[allow(dead_code)]
     fn get_output(&self) -> Vec<String> {
         self.output.lock().unwrap().clone()
     }