use vload3 for half3 or float3 input vector reading,
also check read position to see if it exceed input width
Signed-off-by: Li Peng <peng.li@intel.com>

"custom layers" feature will not work properly in these builds.

use cv::AutoBuffer::data() to get data pointer

- added .data() methods
- added operator[] (int i)
- extend checks support to generic and debug-only cases
- deprecate existed operator* ()