Make the output sort by fragmentation by default.

Also make the usage of "--sort" option consistent with other
perf tools. That is, we support multi keys: "--sort
key1[,key2]...".

 # ./perf kmem --stat caller
 ------------------------------------------------------------------------------
 Callsite                    |Total_alloc/Per | Total_req/Per | Hit  | Frag
 ------------------------------------------------------------------------------
 __netdev_alloc_skb+23       |    5048/1682   |    4564/1521  |     3|   9.588%
 perf_event_alloc.clone.0+0  |    7504/682    |    7128/648   |    11|   5.011%
 tracepoint_add_probe+32e    |     157/31     |     154/30    |     5|   1.911%
 alloc_buffer_head+16        |     456/57     |     448/56    |     8|   1.754%
 radix_tree_preload+51       |     584/292    |     576/288   |     2|   1.370%
 ...

TODO:
- Extract duplicate code in builtin-kmem.c and builtin-sched.c
  into util/sort.c.
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: linux-mm@kvack.org <linux-mm@kvack.org>
LKML-Reference: <4B0B6E72.7010200@cn.fujitsu.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>