Cache the result of countOccupied(gen->large_objects) as gen->n_large_words (#7257)

The program in #7257 was spending 90% of its time counting the live
data in gen->large_objects.  We already avoid doing this for small
objects, but in this example the old generation was full of large
objects (actually pinned ByteStrings).

includes/rts/storage/GC.h

@@ -75,6 +75,7 @@ typedef struct generation_ {
 
     bdescr *       large_objects;	// large objects (doubly linked)
     memcount       n_large_blocks;      // no. of blocks used by large objs
+    memcount       n_large_words;       // no. of words used by large objs
     memcount       n_new_large_words;   // words of new large objects
                                         // (for allocation stats)
 

rts/sm/GC.c

@@ -578,6 +578,7 @@ GarbageCollect (nat collect_gen,
         freeChain(gen->large_objects);
         gen->large_objects  = gen->scavenged_large_objects;
         gen->n_large_blocks = gen->n_scavenged_large_blocks;
+        gen->n_large_words  = countOccupied(gen->large_objects);
         gen->n_new_large_words = 0;
     }
     else // for generations > N
@@ -589,13 +590,15 @@ GarbageCollect (nat collect_gen,
 	for (bd = gen->scavenged_large_objects; bd; bd = next) {
             next = bd->link;
             dbl_link_onto(bd, &gen->large_objects);
-	}
+            gen->n_large_words += bd->free - bd->start;
+        }
         
 	// add the new blocks we promoted during this GC 
 	gen->n_large_blocks += gen->n_scavenged_large_blocks;
     }
 
     ASSERT(countBlocks(gen->large_objects) == gen->n_large_blocks);
+    ASSERT(countOccupied(gen->large_objects) == gen->n_large_words);
 
     gen->scavenged_large_objects = NULL;
     gen->n_scavenged_large_blocks = 0;

rts/sm/Storage.c

@@ -78,6 +78,7 @@ initGeneration (generation *gen, int g)
     gen->n_old_blocks = 0;
     gen->large_objects = NULL;
     gen->n_large_blocks = 0;
+    gen->n_large_words = 0;
     gen->n_new_large_words = 0;
     gen->scavenged_large_objects = NULL;
     gen->n_scavenged_large_blocks = 0;
@@ -951,7 +952,7 @@ W_ countOccupied (bdescr *bd)
 
 W_ genLiveWords (generation *gen)
 {
-    return gen->n_words + countOccupied(gen->large_objects);
+    return gen->n_words + gen->n_large_words;
 }
 
 W_ genLiveBlocks (generation *gen)