Scripts for downloading models in DNN samples

* Initial commit. Utility classes and functions for downloading files

* updated download script

* Support YAML parsing, update download script and configs

* Fix problem with archived files

* fix models.yml

* Move download utilities to more appropriate place

* Fix script description

* Update README

* update utilities for broader range of files

* fix loading with no hashsum provided

* remove unnecessary import

* fix for Python2

* Add usage examples for downloadFile function

* Add more secure cache folder selection

* Remove trailing whitespaces

* Fix indentation

* Update function interface

* Change function for temp dir, change entry name in models.yml

* Update getCacheDirectory function call

* Return python implementation for cache directory selection, use more specific env variable

* Fix whitespace

support SIMD for larger symmetric Bit-exact 16U gaussian blur

* support SIMD for bit-exact 16U symmetric gaussian blur

* use tighter SIMD registers

- detect case with infinite loop and raise NoConv exception
- handle such exception
- add support for case with missing `blobDetector` (image contains Point2f array of candidates)
- add regression test
- undone rectification for "failed" detections too
- drop redirectError() usage

Relates to OpenCV issue #18721

- allow to specify MKL_LIBRARIES through command-line

- due to lifetime restrictions on passed 'features' parameter and its dangling pointers
- dangling pointer issue is fixed for OpenCV 4.5.1+

- allow to load opencv_face_detector with external unpatched protobuf

Fixes OpenCV issue #18721

- NEON / ANDROID_ARM_NEON are toolchain-specific flags
- they are usually not defined for native builds
- let work CPU_BASELINE=DETECT properly

- CAP_PROP_POS_MSEC is not reliable

- don't use OPENCV_ALLOCATOR_STATS_COUNTER_TYPE definition in non C++11 builds
- don't use with MinGW