vaapidecoder shows 1.5x increase in bistream utilization (reported by intel-gpu-top) when decoding interlaced counterpart of the same exact progressive h.264 contents.

Environment:
--------------
kernel:               3.13.0-24-generic #47-Ubuntu x86_64
cpu:                  Intel(R) Atom(TM) CPU  E3845  @ 1.91GHz
gstreamer-vaapi:      (master)6b73746ab6453253ea4bb82170914e0a1892e135 > 0.5.10
libva:                (master)f9741725839ea144e9a6a1827f74503ee39946c3 > 1.4.0
libva-intel-driver:   (master)9a20d6c34cb65e5b85dd16d6c8b3a215c5972b18
gstreamer:            (master)1f6d5d3ff5cb03527acb99b4059788f53d49567e
gst-plugins-base:     (master)e2a693656b1a8513adab5e8f7698758c583c8eed > 1.4.0
gst-plugins-good:     (master)6b69ef24a1e13f0ad7980fe25d954d37133ba746 > 1.4.0
gst-plugins-bad:      (master)8e2964ee599853fff2a7aba9275960b4ae7844fc > 1.4.1
gst-plugins-ugly:     (master)86562f56e208d0d7360a9b5f7bc5cf13836f06a3 > 1.4.0


Steps:
--------------
Used the following pipeline with similar video contents one interlaced and the other progressive. The bitstream utilization as reported by intel-gpu-overlay increases to 1.5x when decoding interlaced contents.

udpsrc address=239.4.4.23 port=10000 name=udpsrc1 ! queue max-size-bytes=0 max-size-buffers=0 max-size-time=10000000000 name=q_tsdemux_1 ! tsdemux name=dmx1 program-number=3 \ 
 dmx1. ! queue max-size-bytes=0 max-size-buffers=0 max-size-time=10000000000 name=q_videoparse_1 ! h264parse ! vaapidecode name=vaapidecode1 ! queue max-size-bytes=0 max-size-buffers=0 max-size-time=10000000000 name=q_videosink_1 ! vaapisink display=x11