python: improve resource.status with many drained ranks

Problem: The ResourceStatus class performs a couple unnecessary idset
set operations for each drain entry in the resource.status response
object. When there are many entries, these set operations perform a
lot of unnecesary work and slow down the _recalcultate() method.

Build the set of drained ranks in the loop (which should be fast),
and delay splitting the set into drained and draining after the loop.

Author: grondo

src/bindings/python/flux/resource/status.py

@@ -112,11 +112,15 @@ def _recalculate(self, include_ranks=None):
             ranks = IDset(drain_ranks)
             if include_ranks is not None:
                 ranks = ranks.intersect(include_ranks)
-            self.drained += ranks - self.allocated
-            self.draining += ranks - self.drained
+            self.drained += ranks
             info = DrainInfo(ranks, entry["timestamp"], entry["reason"])
             self.drain_info.append(info)
 
+        # create the set of draining ranks as the intersection of
+        #  drained and allocated
+        self.draining = self.drained & self.allocated
+        self.drained -= self.draining
+
         # available: all ranks not excluded or drained/draining
         self.avail = self.all - self.get_idset("exclude", "drained", "draining")