feat(orchestrator): Phase 2 engine + tokenizer + tool dispatch

Orchestrator engine:
- engine.rs: unified Mistral Conversations API tool loop that emits
  OrchestratorEvent instead of calling Matrix/gRPC directly
- tool_dispatch.rs: ToolSide routing (client vs server tools)
- Memory loading stubbed (migrates in Phase 4)

Server-side tokenizer:
- tokenizer.rs: HuggingFace tokenizers-rs with Mistral's BPE tokenizer
- count_tokens() for accurate usage metrics
- Loads from local tokenizer.json or falls back to bundled vocab
- Config: mistral.tokenizer_path (optional)

No behavior change — engine is wired but not yet called from
sync.rs or session.rs (Phase 2 continuation).

src/main.rs

@@ -13,6 +13,7 @@ mod orchestrator;
 mod sdk;
 mod sync;
 mod time_context;
+mod tokenizer;
 mod tools;
 
 use std::sync::Arc;
@@ -123,6 +124,13 @@ async fn main() -> anyhow::Result<()> {
     )?;
     let mistral = Arc::new(mistral_client);
 
+    // Initialize tokenizer for accurate token counting
+    let _tokenizer = Arc::new(
+        tokenizer::SolTokenizer::new(config.mistral.tokenizer_path.as_deref())
+            .expect("Failed to initialize tokenizer"),
+    );
+    info!("Tokenizer initialized");
+
     // Build components
     let system_prompt_text = system_prompt.clone();
     let personality = Arc::new(Personality::new(system_prompt));