align allocations and PyGC_Head to 16 bytes on 64-bit platforms

Combination of two upstream commits:
- gh#python/cpython@1b85f4e (3.7.17)
- gh#python/cpython@8766cb7 (3.7.17)

Fixes: bpo#27987
Fixes: rh#1923658
Patch: align-allocations-PyGC_Head.patch

Include/objimpl.h

@@ -255,7 +255,11 @@ typedef union _gc_head {
         union _gc_head *gc_prev;
         Py_ssize_t gc_refs;
     } gc;
-    double dummy;  /* force worst-case alignment */
+    long double dummy;  /* force worst-case alignment */
+    // malloc returns memory block aligned for any built-in types and
+    // long double is the largest standard C type.
+    // On amd64 linux, long double requires 16 byte alignment.
+    // See bpo-27987 for more discussion.
 } PyGC_Head;
 
 extern PyGC_Head *_PyGC_generation0;

Objects/obmalloc.c

@@ -650,8 +650,14 @@ static int running_on_valgrind = -1;
  *
  * You shouldn't change this unless you know what you are doing.
  */
+
+#if SIZEOF_VOID_P > 4
+#define ALIGNMENT              16               /* must be 2^N */
+#define ALIGNMENT_SHIFT         4
+#else
 #define ALIGNMENT               8               /* must be 2^N */
 #define ALIGNMENT_SHIFT         3
+#endif
 
 /* Return the number of bytes in size class I, as a uint. */
 #define INDEX2SIZE(I) (((uint)(I) + 1) << ALIGNMENT_SHIFT)