gh-130030: Fix crash on 32-bit Linux with free threading

The `gc_get_refs` assertion needs to be after we check the alive and
unreachable bits. Otherwise, `ob_tid` may store the actual thread id
instead of the computed `gc_refs`, which may trigger the assertion if
the `ob_tid` looks like a negative value.

Also fix a few type warnings on 32-bit systems.

Author: colesbury

Objects/obmalloc.c

@@ -3297,12 +3297,12 @@ static bool _collect_alloc_stats(
 static void
 py_mimalloc_print_stats(FILE *out)
 {
-    fprintf(out, "Small block threshold = %zd, in %u size classes.\n",
-        MI_SMALL_OBJ_SIZE_MAX, MI_BIN_HUGE);
-    fprintf(out, "Medium block threshold = %zd\n",
-            MI_MEDIUM_OBJ_SIZE_MAX);
-    fprintf(out, "Large object max size = %zd\n",
-            MI_LARGE_OBJ_SIZE_MAX);
+    fprintf(out, "Small block threshold = %zu, in %u size classes.\n",
+        (size_t)MI_SMALL_OBJ_SIZE_MAX, MI_BIN_HUGE);
+    fprintf(out, "Medium block threshold = %zu\n",
+            (size_t)MI_MEDIUM_OBJ_SIZE_MAX);
+    fprintf(out, "Large object max size = %zu\n",
+            (size_t)MI_LARGE_OBJ_SIZE_MAX);
 
     mi_heap_t *heap = mi_heap_get_default();
     struct _alloc_stats stats;

Python/gc_free_threading.c

@@ -564,8 +564,8 @@ typedef struct {
 } gc_span_stack_t;
 
 typedef struct {
-    unsigned int in;
-    unsigned int out;
+    Py_ssize_t in;
+    Py_ssize_t out;
     _PyObjectStack stack;
     gc_span_stack_t spans;
     PyObject *buffer[BUFFER_SIZE];
@@ -574,14 +574,14 @@ typedef struct {
 
 
 // Returns number of entries in buffer
-static inline unsigned int
+static inline Py_ssize_t
 gc_mark_buffer_len(gc_mark_args_t *args)
 {
     return args->in - args->out;
 }
 
 // Returns number of free entry slots in buffer
-static inline unsigned int
+static inline Py_ssize_t
 gc_mark_buffer_avail(gc_mark_args_t *args)
 {
     return BUFFER_SIZE - gc_mark_buffer_len(args);
@@ -1074,14 +1074,14 @@ mark_heap_visitor(const mi_heap_t *heap, const mi_heap_area_t *area,
         return true;
     }
 
-    _PyObject_ASSERT_WITH_MSG(op, gc_get_refs(op) >= 0,
-                                  "refcount is too small");
-
     if (gc_is_alive(op) || !gc_is_unreachable(op)) {
         // Object was already marked as reachable.
         return true;
     }
 
+    _PyObject_ASSERT_WITH_MSG(op, gc_get_refs(op) >= 0,
+                                  "refcount is too small");
+
     // GH-129236: If we've seen an active frame without a valid stack pointer,
     // then we can't collect objects with deferred references because we may
     // have missed some reference to the object on the stack. In that case,