refactor(commit-ai): remove local models and simplify AI flow

Remove local on-device model support and related IPC commands,
consolidating commit-generation to cloud providers and simplifying the
AI crate surface. Local-specific types, generation profiles, caching,
and the local prompt builder were removed while message sanitization and
diff-echo detection were preserved. Also harden repository handling and
runtime: unborn HEADs are handled gracefully so empty repos still report
files, Git LFS sync is skipped for repositories without commits, and
tokio runtime features were enabled.

- Remove local model engine, load/status commands, and local profile code
- Handle unborn HEAD and skip LFS sync for repos without commits
- Enable tokio runtime features and route AI generation to cloud only

src-tauri/crates/commit_ai/Cargo.toml

@@ -5,8 +5,6 @@ version = "0.1.0"
 edition = "2024"
 
 [dependencies]
-mistralrs = "0.8"
-tokio = { version = "1", features = ["sync"] }
 serde = { version = "1", features = ["derive"] }
 serde_json = "1"
 reqwest = { version = "0.12", default-features = false, features = ["json", "rustls-tls"] }