dagop: comment why revancestors() doesn't heapify input revs at once

I wondered why we're doing this complicated stuff without noticing the input
revs may be iterated lazily in descending order. c1f666e27345 showed why.

mercurial/dagop.py

@@ -26,6 +26,9 b' def _genrevancestors(repo, revs, followf'
     else:
         cut = None
     cl = repo.changelog
+
+    # load input revs lazily to heap so earlier revisions can be yielded
+    # without fully computing the input revs
     revs.sort(reverse=True)
     irevs = iter(revs)
     h = []