Add batching to arena bins.

This adds a fast-path for threads freeing a small number of allocations to
bins which are not their "home-base" and which encounter lock contention in
attempting to do so. In producer-consumer workflows, such small lock hold times
can cause lock convoying that greatly increases overall bin mutex contention.

test/include/test/fork.h

@@ -0,0 +1,32 @@
+#ifndef JEMALLOC_TEST_FORK_H
+#define JEMALLOC_TEST_FORK_H
+
+#ifndef _WIN32
+
+#include <sys/wait.h>
+
+static inline void
+fork_wait_for_child_exit(int pid) {
+	int status;
+	while (true) {
+		if (waitpid(pid, &status, 0) == -1) {
+			test_fail("Unexpected waitpid() failure.");
+		}
+		if (WIFSIGNALED(status)) {
+			test_fail("Unexpected child termination due to "
+			    "signal %d", WTERMSIG(status));
+			break;
+		}
+		if (WIFEXITED(status)) {
+			if (WEXITSTATUS(status) != 0) {
+				test_fail("Unexpected child exit value %d",
+				    WEXITSTATUS(status));
+			}
+			break;
+		}
+	}
+}
+
+#endif
+
+#endif /* JEMALLOC_TEST_FORK_H */

test/include/test/jemalloc_test.h.in

@@ -1,3 +1,6 @@
+#ifndef JEMALLOC_TEST_H
+#define JEMALLOC_TEST_H
+
 #ifdef __cplusplus
 extern "C" {
 #endif
@@ -172,3 +175,5 @@ extern "C" {
 #ifdef __cplusplus
 }
 #endif
+
+#endif