feat: add generation metrics to logs (think, rounds, tokens)

Log think flag, round count, prompt/output token counts per generation.
Change log category from 'usage' to 'generation' for clean MCP filtering.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

src/fabledassistant/services/llm.py

@@ -149,6 +149,9 @@ class ChatChunk:
     type: Literal["content", "thinking", "tool_calls", "done"]
     content: str = ""
     tool_calls: list[dict] | None = None
+    # Token counts from the Ollama done event (only set on type="done")
+    prompt_tokens: int | None = None
+    output_tokens: int | None = None
 
 
 async def stream_chat_with_tools(
@@ -221,7 +224,11 @@ async def stream_chat_with_tools(
                         yield ChatChunk(type="tool_calls", tool_calls=accumulated_tool_calls)
                     else:
                         logger.debug("Ollama done with no tool calls")
-                    yield ChatChunk(type="done")
+                    yield ChatChunk(
+                        type="done",
+                        prompt_tokens=data.get("prompt_eval_count"),
+                        output_tokens=data.get("eval_count"),
+                    )
                     break