Possible #1872 issue fix

I'm going to update this commit if it holds true.
I'd like some additional testing.

@dipinhora @sylvanc can you test #1872 for this.
@Cloven can you see if this fixes the issue for you.

@agarman @Theodus I don't think this will fix #1483 but
can you check and see as there is a chance it might.

src/libponyrt/ds/hash.c

@@ -114,7 +114,6 @@ static void resize(hashmap_t* map, cmp_fn cmp, alloc_fn alloc,
 static size_t optimize_item(hashmap_t* map, alloc_fn alloc,
   free_size_fn fr, cmp_fn cmp, size_t old_index)
 {
-
   size_t mask = map->size - 1;
 
   size_t h = map->buckets[old_index].hash;
@@ -133,17 +132,26 @@ static size_t optimize_item(hashmap_t* map, alloc_fn alloc,
     ib_index = index >> HASHMAP_BITMAP_TYPE_BITS;
     ib_offset = index & HASHMAP_BITMAP_TYPE_MASK;
 
-    // don't need to check probe counts for filled buckets because
-    // earlier items are guaranteed to have a lower probe count
-    // than us and we cannot displace them
-    // found an earlier empty bucket so move item
     if((map->item_bitmap[ib_index] & ((bitmap_t)1 << ib_offset)) == 0)
     {
       ponyint_hashmap_clearindex(map, old_index);
       ponyint_hashmap_putindex(map, entry, h, cmp, alloc, fr, index);
       return 1;
     }
-
+    else
+    {
+      size_t item_probe_length =
+        get_probe_length(map, h, index, mask);
+      size_t there_probe_length =
+        get_probe_length(map, map->buckets[index].hash, index, mask);
+
+      if (item_probe_length > there_probe_length)
+      {
+        ponyint_hashmap_clearindex(map, old_index);
+        ponyint_hashmap_putindex(map, entry, h, cmp, alloc, fr, index);
+        return 1;
+      }
+    }
     // find next bucket index
     index = (index + 1) & mask;
   }