Improve readability of InputArray type arg-s in func-s signature (Feature #2518)

[from ECCV-2012 / Mobile OpenCV tutorial / wishlist brainstorm]
Some of OpenCV functions ( e.g. cv::add() ) accepts input arg-s of almost any types, so InputArray works like a charm here.
But there are some others that expect InputArray-s of particular type(s) only, and use of InputArray type for input arg-s here complicates their usage. Examples:

• findHomography: the InputArray srcPoints, InputArray dstPoints are expected to be 2D points vector, i.e. std::vector<Point2{i,f,d}>, Mat 1xN or Nx1 CV_{32S,32F,64F}C2
• convexityDefects: the InputArray _points is expected to be std::vector<Point2i>, Mat 1xN or Nx1 CV_32SC2
• calcOpticalFlowPyrLK:
  • InputArray prevImg, InputArray nextImg are expected to be gray-scale images (8UC1) or pyramid (std::vector<Mat>)
  • InputArray prevPts, CV_OUT InputOutputArray nextPts are expected to be Point2f vector
    and etc.
    Even for cv::add() the InputArray src2 can be not only image but cv::Scalar as well!

The suggestion is to define a set of typedef-s that will help to understand what is expected and use them in signatures.

1. randu
2. randn
3. inRange
4. goodFeaturesToTrack
5. findFundamentalMat
6. cornerSubPix
7. minEnclosingCircle
8. findHomography
9. solvePnP
10. distCoeffs
11. solvePnPRansac
12. distCoeffs
13. calcOpticalFlowPyrLK
14. fitEllipse
15. fillPoly
16. polylines
17. fillConvexPoly
18. boundingRect
19. approxPolyDP
20. arcLength
21. pointPolygonTest
22. minAreaRect
23. getAffineTransform
24. hconcat
25. vconcat
26. undistortPoints
27. checkRange
28. meanStdDev
29. drawContours
30. findContours
31. convexityDefects
32. isContourConvex
33. convexHull
34. projectPoints
35. distCoeffs
36. initCameraMatrix2D
37. findChessboardCorners
38. drawChessboardCorners