fix(haivemind): start engine after MainScreen mounts

Upstream haivemind_tui.py started the engine synchronously in HaivemindApp.on_mount while MainScreen mounts asynchronously, causing ProbeStart handler to query #chat-log on the default screen (NoMatches). Move the start_run trigger into MainScreen.on_mount so it runs only once the screen is active.

packages/haivemind/fix-chat-log-race.patch

@@ -0,0 +1,28 @@
+--- a/haivemind_tui.py
++++ b/haivemind_tui.py
+@@ -1195,6 +1195,14 @@
+         yield TheatrePanel(id="theatre-container")
+         yield Footer()
+ 
++    def on_mount(self) -> None:
++        app = self.app
++        if (
++            getattr(app, "_prompt", None) is not None
++            and getattr(app, "_config", None) is not None
++        ):
++            app.start_run(app._prompt, app._config)
++
+ 
+ class HaivemindApp(App):
+     CSS = CSS
+@@ -1224,9 +1232,7 @@
+     def on_mount(self) -> None:
+         self.push_screen(MainScreen())
+         self.set_interval(0.15, self._tick_spinner)
+-        if self._prompt is not None and self._config is not None:
+-            self.start_run(self._prompt, self._config)
+-        else:
++        if self._prompt is None or self._config is None:
+             self.push_screen(ConfigScreen())
+ 
+     def start_run(self, prompt: str, config: HaivemindConfig) -> None:

packages/haivemind/package.nix

@@ -24,6 +24,8 @@ python3Packages.buildPythonApplication rec {
     textual
   ];
 
+  patches = [ ./fix-chat-log-race.patch ];
+
   doCheck = false;
   pythonImportsCheck = [
     "haivemind"