refactor(llm-api): implement clean architecture following scraper pattern

Split monolithic 1012-line main.rs into layered hexagonal architecture:
- Domain: entity types and LlmError enum
- Application: prompt building, sampler construction, tool call parsing
- Infrastructure: LlamaEngine wrapping llama-cpp-2 with isolated unsafe transmute
- Presentation: Axum handlers, middleware (auth), error chain, router
- Config: type-safe AppConfig with LazyLock
- Bootstrap: Application struct with build() + run()

Resolves build_sampler/build_sampler_params duplication.
Adds simple web chat UI at GET /.

Co-Authored-By: Claude Code <noreply@anthropic.com>

src/lib.rs

@@ -0,0 +1,27 @@
+// Library root — clean architecture module structure
+
+// ============================================================================
+// Domain Layer — pure business logic, no framework dependencies
+// ============================================================================
+pub mod domain;
+
+// ============================================================================
+// Application Layer — use cases / business orchestration
+// ============================================================================
+pub mod application;
+
+// ============================================================================
+// Infrastructure Layer — implements domain ports
+// ============================================================================
+pub mod infrastructure;
+
+// ============================================================================
+// Presentation Layer — Axum handlers, middleware, state, error
+// ============================================================================
+pub mod presentation;
+
+// ============================================================================
+// Core Framework & Bootstrap
+// ============================================================================
+pub mod bootstrap;
+pub mod config;