debug_is_empty: don't dealloc() free space

We know that these are small objects and that they have already had
their cache friendly offsets applied, so jump in to the dealloc path
futher down, below the stats calls.

Split small_dealloc_offseted into a wrapper around the core to avoid
inlining too much into the debug function.

While here, move the handling of the Remote objects into its own block
so that `p` is out of scope thereafter.

Fixes https://github.com/microsoft/snmalloc/issues/98

src/mem/alloc.h

@@ -895,13 +895,15 @@ namespace snmalloc
       };
 
       // Destroy the message queue so that it has no stub message.
-      Remote* p = message_queue().destroy();
-
-      while (p != nullptr)
       {
-        Remote* n = p->non_atomic_next;
-        handle_dealloc_remote(p);
-        p = n;
+        Remote* p = message_queue().destroy();
+
+        while (p != nullptr)
+        {
+          Remote* n = p->non_atomic_next;
+          handle_dealloc_remote(p);
+          p = n;
+        }
       }
 
       for (size_t i = 0; i < NUM_SMALL_CLASSES; i++)
@@ -911,7 +913,10 @@ namespace snmalloc
         while (prev != nullptr)
         {
           auto n = Metaslab::follow_next(prev);
-          dealloc(remove_cache_friendly_offset(prev, i));
+
+          Superslab* super = Superslab::get(prev);
+          small_dealloc_offseted_inner(super, prev, i);
+
           prev = n;
         }
 
@@ -1222,6 +1227,12 @@ namespace snmalloc
       MEASURE_TIME(small_dealloc, 4, 16);
       stats().sizeclass_dealloc(sizeclass);
 
+      small_dealloc_offseted_inner(super, p, sizeclass);
+    }
+
+    SNMALLOC_FAST_PATH void small_dealloc_offseted_inner(
+      Superslab* super, void* p, sizeclass_t sizeclass)
+    {
       Slab* slab = Slab::get(p);
       if (likely(slab->dealloc_fast(super, p)))
         return;