sfm: Fix declarations to make Python bindings work

[emilepaul]

See issue #636

I ran into the same issue and would like to discuss/propose the remedy in this PR which allowed me to use the SFM module from Python.

Note 1: I had to add a CV_WRAP for SFMLibmvEuclideanReconstruction::create, just running the constructor produced a NULL self in pyopencv_generated_types_content.h when ->run() was called.

Note 2: Most of the other problems were solved by replacing OutputArray to OutputArrayOfArrays (distinction doesn't matter to the C++ type system but does matter to the Python bindings generator) to prevent GetMatRef(i) from throwing an assert on the OutputArray not being of a STD_VECTOR_MAT kind in all routines returning Ts and points3d.

I don't know if using a vector is the best choice for Ts and points3d output, I'd say a regular Mat with 3 columns would produce nicer Numpy arrays and something that serializes nicer to cv::FileStorage. But I didn't want to change any semantics.

Note 3: I also had to shorten SFMLibmvEuclideanReconstruction to SFMLibmvEuclideanRecons in order to avoid the Python bindings generator shortening/garbling it to SFMLibmvEuclideanReruction in pyopencv_generated_types_content.h (could also be that something overenthousiastically strips 'const' in the wrong way, I'm no expert on the bindings generators).

• I agree to contribute to the project under Apache 2 License.
• To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
• The PR is proposed to the proper branch
• There is a reference to the original bug report and related work
• There is accuracy test, performance test and test data in opencv_extra repository, if applicable
  Patch to opencv_extra has the same branch name.
• The feature is well documented and sample code can be built with the project CMake