Function load_to_shared 

pub fn load_to_shared<T: RudaPrimitive>(
    input: &Array<T>,
    output: &mut SharedMemory<T>,
    offset: usize,
    valid_items: usize,
    padding: T,
    threads: usize,
    tile_items: usize,
)

Cooperatively load a tile into shared memory; every thread participates.