Optimize Valgrind integration.

Forcefully disable tcache if running inside Valgrind, and remove
Valgrind calls in tcache-specific code.

Restructure Valgrind-related code to move most Valgrind calls out of the
fast path functions.

Take advantage of static knowledge to elide some branches in
JEMALLOC_VALGRIND_REALLOC().

src/valgrind.c

@@ -0,0 +1,34 @@
+#include "jemalloc/internal/jemalloc_internal.h"
+#ifndef JEMALLOC_VALGRIND
+#  error "This source file is for Valgrind integration."
+#endif
+
+#include <valgrind/memcheck.h>
+
+void
+valgrind_make_mem_noaccess(void *ptr, size_t usize)
+{
+
+	VALGRIND_MAKE_MEM_NOACCESS(ptr, usize);
+}
+
+void
+valgrind_make_mem_undefined(void *ptr, size_t usize)
+{
+
+	VALGRIND_MAKE_MEM_UNDEFINED(ptr, usize);
+}
+
+void
+valgrind_make_mem_defined(void *ptr, size_t usize)
+{
+
+	VALGRIND_MAKE_MEM_DEFINED(ptr, usize);
+}
+
+void
+valgrind_freelike_block(void *ptr, size_t usize)
+{
+
+	VALGRIND_FREELIKE_BLOCK(ptr, usize);
+}