allows disabling automatic conversion to RGB for certain formats. If
conversion is disabled the returned image just points to the underlying
buffer, so no memcpy is performed.

Note that we do not check image size in retrieve frame any more as it
can not possibly while the device is streaming.
Furthermore this code was disabled altogether by the wrong ifdef in the
previous commit.

in the case of HAVE_CAMV4L2 && !HAVE_CAMV4L there was no body for the if
statement.

they are upstream since at least linux 2.6.8 (10 Years)
http://lxr.oss.org.cn/ident?v=2.6.8&i=V4L2_PIX_FMT_SBGGR8

do not use a custom enum instead of the V4L2 fourcc defines for palette.
This way we can easily implement CAP_PROP_FOURCC and CAP_PROP_MODE.

use logic similar to cap_libv4l: replace icvSetVideoSize by v4l2_reset
as it was not used for V4L1, the actual frame format is negotiated in
try_palette_v4l2 and the stream has to restarted anyway.

now icvGetPropertyCAM_V4L behaves the same as in cap_libv4l. This also
fixes passing boolean values.

also refactor property range handling and opencv property to V4L2
translation.

Catched on Ubuntu 12.04 with FFMPEG:
  codec:                     YES (ver 53.35.0)
  format:                    YES (ver 53.21.1)
  util:                      YES (ver 51.22.3)
  swscale:                   YES (ver 2.1.0)
  resample:                  NO
  gentoo-style:              YES

it is new'ed but never deleted. However it does not need to be global or
heap allocated in the first place.

Corrected typo and unused imports

OPENCV_IPP env var now allows to select IPP architecture level for IPP9+;
IPP initialization logic was unified across modules;