More efficient Python slicing

Call reserve for container types that support it to avoid repeated
memory reallocations for new slices or slices that grow in size.

Lib/python/std_unordered_multimap.i

@@ -16,6 +16,13 @@
       }
     }
 
+    template <class K, class T>
+    struct traits_reserve<std::unordered_multimap<K,T> >  {
+      static void reserve(std::unordered_multimap<K,T> &seq, typename std::unordered_multimap<K,T>::size_type n) {
+        seq.reserve(n);
+      }
+    };
+
     template <class K, class T>
     struct traits_asptr<std::unordered_multimap<K,T> >  {
       typedef std::unordered_multimap<K,T> unordered_multimap_type;