Add version checks around uses of AVCodecID and av_opt_set, since these aren't
defined for older versions of ffmpeg.

Add a call to avcodec_get_context_defaults3, this sets per-codec default values
and in the case of libx264 will stop VideoWriter::open (with fourcc=x264) from
failing due to libx264 complaining about broken ffmpeg default settings.

Set some additional libx264 encoder parameters.

    Update find_obj.py:
    Added additional error checking for the loaded files and modified the detector check to be more pythonic (is versus ==).

    Update common.py:
    Fixed a typo (contais -> contains)

fix condition for sample index in assert

And a potential problem in remap

In class Mat, "static MatExpr Mat::zeros(int ndims, const int* sz, int
type)" and "static MatExpr Mat::ones(int ndims, const int* sz, int
type)" are declared but never implemented. That means we can see their
manuals from
"http://docs.opencv.org/modules/core/doc/basic_structures.html" but we
can't use them. Here I tried to finish their implementation.
I have also changed MatOp_Initializer::assign to make it support
multi-dimension Mat.
Test cases are added in test_math.cpp as well.