Please describe the problem:
In gst_multipart_mux_collected, caps are allocated but not unreffed after setting them on the pad (using gst_pad_set_caps), which introduces a memory leak of the caps.

Steps to reproduce:
Custom code, haven't been able to repeat with valgrind + gst-launch.

Actual results:
=14985== 191 (32 direct, 159 indirect) bytes in 1 blocks are definitely lost in loss record 3,795 of 3,920
==14985==    at 0x401C38B: malloc (vg_replace_malloc.c:149)
==14985==    by 0x4375915: g_malloc (gmem.c:131)
==14985==    by 0x438A752: g_slice_alloc (gslice.c:824)
==14985==    by 0x41482A1: gst_caps_new_empty (gstcaps.c:142)
==14985==    by 0x4148470: gst_caps_new_simple (gstcaps.c:194)
==14985==    by 0x5CEA7D5: gst_multipart_mux_collected (multipartmux.c:485)
==14985==    by 0x4069C3B: gst_collect_pads_check_collected (gstcollectpads.c:967)
==14985==    by 0x406B47B: gst_collect_pads_chain (gstcollectpads.c:1176)
==14985==    by 0x41AA0DB: gst_pad_chain_unchecked (gstpad.c:3518)
==14985==    by 0x41AB26D: gst_pad_push (gstpad.c:3686)
==14985==    by 0x405414B: gst_base_src_loop (gstbasesrc.c:2111)
==14985==    by 0x41EBB1E: gst_task_func (gsttask.c:192)
==14985==    by 0x4393306: g_thread_pool_thread_proxy (gthreadpool.c:265)
==14985==    by 0x439177E: g_thread_create_proxy (gthread.c:635)
==14985==    by 0x4402C50: pthread_start_thread (manager.c:310)
==14985==    by 0x4509419: clone (clone.S:119)


Expected results:
No leak.

Does this happen every time?
Yes

Other information:
Patch suggestion:

-    if (!gst_pad_set_caps (mux->srcpad, newcaps))
+    if (!gst_pad_set_caps (mux->srcpad, newcaps)) {
+      gst_caps_unref (newcaps);
       goto nego_error;
+    }
 
+    gst_caps_unref (newcaps);
     mux->negotiated = TRUE;
   }