Fix inplace update buffer self-deadlock.

A CacheInvalidateHeapTuple* callee might call
CatalogCacheInitializeCache(), which needs a relcache entry.  Acquiring
a valid relcache entry might scan pg_class.  Hence, to prevent
undetected LWLock self-deadlock, CacheInvalidateHeapTuple* callers must
not hold BUFFER_LOCK_EXCLUSIVE on buffers of pg_class.  Move the
CacheInvalidateHeapTupleInplace() before the BUFFER_LOCK_EXCLUSIVE.  No
back-patch, since I've reverted commit
243e9b40f1 from non-master branches.

Reported by Alexander Lakhin.  Reviewed by Alexander Lakhin.

Discussion: https://postgr.es/m/10ec0bc3-5933-1189-6bb8-5dec4114558e@gmail.com

src/backend/utils/cache/inval.c

@@ -1201,6 +1201,18 @@ AtInplace_Inval(void)
 	inplaceInvalInfo = NULL;
 }
 
+/*
+ * ForgetInplace_Inval
+ *		Alternative to PreInplace_Inval()+AtInplace_Inval(): discard queued-up
+ *		invalidations.  This lets inplace update enumerate invalidations
+ *		optimistically, before locking the buffer.
+ */
+void
+ForgetInplace_Inval(void)
+{
+	inplaceInvalInfo = NULL;
+}
+
 /*
  * AtEOSubXact_Inval
  *		Process queued-up invalidation messages at end of subtransaction.