Given an array of bytes, representing N elements of a type with S bytes, the filter rearrange the bytes from
[1_1, 1_2, ..., 1_S, 2_1, 2_2, ..., 2_S, ..., N_1, N_2, ..., N_S] to
[1_1, 2_1, ..., N_1, 1_2, 2_2, ..., N_2, ..., 1_S, 2_S, ..., N_S],
where i_j is the j-th byte of the i-th element.
Truncation precision filter for floating point data.
This filter reduces the precision of floating point numbers by truncating the least
significant bits.
This filter is only supported for floating point types (e.g., f32, f64). This can not
be enforced by the library, there it is only checked that the typesize is 4 or 8 bytes.