Also, added the missing include guard to perf_precomp.hpp.
This should fix the build.

Looks like we don't actually use them; when creating precompiled headers
with Visual C++, a dummy .cpp is created at build time.

ReST directive was used. Also fixed some other ReST directives that were
not correct and removed some warnings during buildbot checks.

Add documentation.

Most codes are ported from AMD's Bolt library.
Four methods are implemented:

SORT_BITONIC,   // only support power-of-2 buffer size
SORT_SELECTION, // cannot sort duplicate keys
SORT_MERGE,
SORT_RADIX      // only support signed int/float keys

- set init value for `numsdev` to prevent use of uninitialized value
- stop use of 'workdir' and files from samples
- forcing use of 'opencv_extra' instead
Note: set OPENCV_TEST_DATA_PATH to full path to 'opencv_extra/testdata' (gitolite@code.opencv.org:opencv_extra.git) before running the test!

Previously the function requires Info::Impl implicitly. A call will create
new Context this function which is not intended.
The properties are now moved to global scope to fix this issue.