[BUG] cvcuda::Remap does not output the correct result on orin, and x86 runs coredump

[Xiao-Hu-Z]

src_img = nvcv::Tensor(n_img_, {img_w_, img_h_}, nvcv::FMT_RGB8);
dst_img = nvcv::Tensor(n_img_, {img_w_, img_h_}, nvcv::FMT_RGB8);

bool   alignCorners{true};
float4 borderValue{0, 0, 0, 0};
cvcuda::Remap remap_op;
remap_op(stream_, src_img, dst_img, map_tensor_, NVCV_INTERP_NEAREST, 
    NVCV_INTERP_NEAREST, NVCV_REMAP_ABSOLUTE_NORMALIZED, static_cast<int8_t>(alignCorners),
    NVCV_BORDER_CONSTANT, borderValue);

[shiremathNV]

@Xiao-Hu-Z - can you share what is the output that you are receiving at your end? Also, what version of CV-CUDA are you currently using?