SAFETY-96 review: keep the refusal count over a refused write; test a fast board

F1: resetting the count on each refused write hid a battery that really
refuses behind a link refusing 1 write in 3 (never learned; learned at
358 s with the count kept). F2: a board at 0 W before its link is stale
learned no refusal untested (mutation R7 survived); now a 1 s-lag case.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

goodwe_controller/app/main.py

@@ -1402,7 +1402,8 @@ class Controller:
             if inv.rejected_since is not None:
                 # SAFETY-96: its writes are refused, so its board may have
                 # gone to 0 W on its own; that is no refusal by the battery.
-                inv.ref_count = 0
+                # The count is kept, as across a gap, so a link that refuses
+                # now and then does not hide a battery that does refuse.
                 continue
             sign = self._gave_nothing(inv)
             if sign: