Optimize Neon implementation of vpx_int_pro_col

Use widening pairwise addition instructions to halve the number of
additions required.

Change-Id: I0307a3b65e50d2b1ae582938bc5df9c2b21df734
1 file changed