tested with NDK 17b (17.1.4828580)

Some `CV_Assert`s changed to `CV_Assert_N`s according to https://github.com/opencv/opencv/issues/12304

* Support Mask-RCNN from TensorFlow

* Fix a sample

When the fullAffine parameter is set to false, the estimateRigidTransform function maybe return empty, then the _localAffineEstimate function will be called, but the bug in it will result in incorrect results.

The while loop would run only once making it useless and leading
to confusion.

So, remove the unwanted while loop and just keep an infinite for
loop.

Intrinsics must be effective, so don't declare FP16 type/operations if there is no native support.

- CV_FP16: supports load/store into/from float32
- CV_SIMD_FP16: declares FP16 types and native FP16 operations