phases: avoid N² behavior in `advanceboundary`

We allowed duplicated entries in the deque, which each entry could potentially
insert all its ancestors. So advancing boundary for the full repository would
mean each revision would walk all its ancestors, resulting in O(N²) iteration.
For repository of any decent size, N² is quickly insane.

We introduce a simple set to avoid this and get back to reasonable performance.

mercurial/phases.py

@@ -703,6 +703,7 b' class phasecache:'
             return set()
 
         # search for affected high phase changesets and roots
+        seen = set(new_revs)
         push = heapq.heappush
         pop = heapq.heappop
         parents = cl.parentrevs
@@ -735,9 +736,11 b' class phasecache:'
                 # higher phases
                 delroots.add(current)
             # schedule a walk down if needed
-            if p1_phase > targetphase:
+            if p1_phase > targetphase and p1 not in seen:
+                seen.add(p1)
                 push(revs, -p1)
-            if p2_phase > targetphase:
+            if p2_phase > targetphase and p2 not in seen:
+                seen.add(p2)
                 push(revs, -p2)
             if p1_phase < targetphase and p2_phase < targetphase:
                 new_target_roots.add(current)