Fix small chunk finaliser premature re-use bug

Prior to this commit, if a small chunk was partially used after GC
and had finalisers to be run for the freed slots, the execution of
the finaliser could end up reusing the slots for which finalisers
were being run. If this occurred, it would lead to unpredictable
results (a segfault in my case).

This commit changes the order of operations to ensure that all the
finalisers are run prior to the chunk being added back to the
available list ensuring that the execution of the finaliser cannot
allocate slots that are still to be finalised.

src/libponyrt/mem/heap.c

@@ -184,11 +184,15 @@ static size_t sweep_small(chunk_t* chunk, chunk_t** avail, chunk_t** full,
     } else {
       used += sizeof(block_t) -
         (__pony_popcount(chunk->slots) * size);
-      chunk->next = *avail;
-      *avail = chunk;
 
       // run finalisers for freed slots
       final_small_freed(chunk);
+
+      // make chunk available for allocations only after finalisers have been
+      // run to prevent premature reuse of memory slots by an allocation
+      // required for finaliser execution
+      chunk->next = *avail;
+      *avail = chunk;
     }
 
     chunk = next;