Bisect the scheduler queue in place instead of slicing it (#191)

Splicing a watcher into a flushing queue sliced the tail of
_queue_indices just to bisect it, copying O(n) entries on every
mid-flush enqueue and then correcting the returned index by hand.
Passing the lo bound to bisect searches the same range in place.


Claude-Session: https://claude.ai/code/session_01UyALuqgF1ZZ3Lj88FzwGVc

Co-authored-by: Claude <noreply@anthropic.com>

Author: Korijn

observ/scheduler.py

@@ -206,8 +206,7 @@ def queue(self, watcher: Watcher[Any]) -> None:
             # If already past its id, it will be run next immediately.
             # Last part of the queue should stay ordered, in order to
             # properly make use of bisect and avoid deadlocks
-            i = bisect(self._queue_indices[self.index + 1 :], watcher.id)
-            i += self.index + 1
+            i = bisect(self._queue_indices, watcher.id, self.index + 1)
             self._queue.insert(i, watcher)
             self._queue_indices.insert(i, watcher.id)