The MD5 sum SEI message is capable of emulating startcodes.  Apply a
temporary fix to allow this to be done before HM-3.0.

This does not fix the lack of emulation-prevention on SPS/PPS NALunits
and this requires fixing more generally.

Reviewed-by: David Flynn <davidf@rd.bbc.co.uk>
[added extra trivial comment]

From: Frank Bossen <bossen@docomolabs-usa.com>

From: David Flynn <davidf@rd.bbc.co.uk>

On the Encoder:
  --SEIpictureDigest=1 (default) calculates and writes per-picture MD5
                       picture_digest SEI messages
  --SEIpictureDigest=0 does not calculate MD5s or write the SEI messages

On the Decoder:
  --SEIpictureDigest=1 (default) calculates per-picture MD5sums and
                       compares against any picture_digest SEI message
  --SEIpictureDigest=0 does not calculate MD5s or interpret picture_digest
                       SEI message

From: David Flynn <davidf@rd.bbc.co.uk>

This bug seems to just cause valgrind errors (and possibly some wasted
trivial computations) and doesn't seem to impact anything else.

From: David Flynn <davidf@rd.bbc.co.uk>

This also corrects the status line to show the correct information even
when ENABLE_IBDI=1

From: David Flynn <davidf@rd.bbc.co.uk>

From: David Flynn <davidf@rd.bbc.co.uk>

Don't read past the end of bNeighborFlags array.

From: David Flynn <davidf@rd.bbc.co.uk>

This option allows n frames will be skipped at input.  It will succeed
if the input file is a regular file or a pipe.

From: David Flynn <davidf@rd.bbc.co.uk>

don't import std namespace in TVideoIOBits.h

Provided by: Xiang Li (Mediatek)

from CW Hsu <cw.hsu@mediatek.com>

NDEBUG is a preprocessor macro that (amongst other things) causes
assert(0) to become a no-op.  The premise being that assert(x) offers
such a strong guarantee that x is true, there is no need to test for
it in release builds.  Unfortunately, in the this software, the
guarantee isn't always very strong, with assert sometimes being used
to replace formal error handling.

Furthermore, due to the slow speed of the software at the best of times,
it doesn't make sense to have build configurations along the lines of
 either (debug): -O0 -g
 or   (release): -O3 -NDEBUG
Rather, what is actually desired is "debugging with and without
optimizations".

From: David Flynn <davidf@rd.bbc.co.uk>

A quick search reveals that this has never been implemented, remove the
option for now, it can be put back later.

From: David Flynn <davidf@rd.bbc.co.uk>

From: David Flynn <davidf@rd.bbc.co.uk>

From: David Flynn <davidf@rd.bbc.co.uk>

Encoder and Decoder have been modified to print a list of unhandled
arguments.

From: David Flynn <davidf@rd.bbc.co.uk>

When scanning argv[] for options:
 - the argument '--' will stop any further processing.
 - the argument '-' is not treated as an option.

From: David Flynn <davidf@rd.bbc.co.uk>

From: David Flynn <davidf@rd.bbc.co.uk>

From: David Flynn <davidf@rd.bbc.co.uk>