Observed during launch of 32/64-bit applications on Windows.
Added '32-bit' prefix for 32-bit OpenCL devices. No prefix on 64-bit configurations.

- don't write into testdata directory
- check matrices instead of result images

This reverts commit 5384a220.

'WITH_' variables is intended to enable CMake scripts with some autodetection logic.
'WITH_' can be off, but components is really enabled via command-line options
with proper variables setup (including 'HAVE_').

Python executable is not used to generate code for iOS builds

Sometimes it tries to use host '/usr/include' directory and garbages
compiler flags with invalid '-I /usr/include'

- Use EXPECT_TRUE instead of EXPECT_EQ for comparing NULL in xlc
- Added support for int64 to vec_promote in xlc, clang
- Fixed v_rotate_right in xlc

- 3rdparty/libjasper/jas_getopt.c:129:25: warning: this function declaration is not a prototype [-Wstrict-prototypes]
- clang: warning: optimization flag '-fomit-frame-pointer' is not supported for target 'armv7' [-Wignored-optimization-argument]

Allow to use 3 points in SolvePnP if SOLVEPNP_ITERATIVE and useExtrinsicGuess==true. Add bibtex citations for P3P. Update SolvPnP tests.

This PR fixes incorrect division by zero handling in template
specialization of `Div_SIMD` for type `double`.

If there are no OpenCL/UMat methods calls from application.

OpenCL subsystem is initialized:
- haveOpenCL() is called from application
- useOpenCL() is called from application
- access to OpenCL allocator: UMat is created (empty UMat is ignored) or UMat <-> Mat conversions are called

Don't call OpenCL functions if OPENCV_OPENCL_RUNTIME=disabled
(independent from OpenCL linkage type)

Universal Intrinsics: aligned v_pack behavior on different platforms, fixed 64-bit register on ARM, added more saturate_cast variants