Use a __thread cache for the GC's thread-local ThreadBlockCache

__thread seems quite a bit faster than pthread_get_specific, so
if we give up on having multiple Heap objects, then we can store
a reference to the current thread's ThreadBlockCache in a static
__thread variable.  It looks like this ends up mattering (5% average
speedup) since SmallArena::_alloc() is so hot

src/gc/heap.cpp

@@ -463,7 +463,9 @@ GCAllocation* SmallArena::_alloc(size_t rounded_size, int bucket_idx) {
     Block** free_head = &heads[bucket_idx];
     Block** full_head = &full_heads[bucket_idx];
 
-    ThreadBlockCache* cache = thread_caches.get();
+    static __thread ThreadBlockCache* cache = NULL;
+    if (!cache)
+        cache = thread_caches.get();
 
     Block** cache_head = &cache->cache_free_heads[bucket_idx];
 

src/gc/heap.h

@@ -110,7 +110,14 @@ static constexpr size_t NUM_BUCKETS = sizeof(sizes) / sizeof(sizes[0]);
 
 class SmallArena : public Arena<SMALL_ARENA_START, ARENA_SIZE> {
 public:
-    SmallArena(Heap* heap) : Arena(), heap(heap), thread_caches(heap, this) {}
+    SmallArena(Heap* heap) : Arena(), heap(heap), thread_caches(heap, this) {
+#ifndef NDEBUG
+        // Various things will crash if we instantiate multiple Heaps/Arenas
+        static bool already_created = false;
+        assert(!already_created);
+        already_created = true;
+#endif
+    }
 
     GCAllocation* __attribute__((__malloc__)) alloc(size_t bytes);
     GCAllocation* realloc(GCAllocation* alloc, size_t bytes);