Explicitly change MV precision
In current VTM code, MV precision is changed via shift, which causes confusion sometimes. In this request, the precision change is explicitly specified from source precision to destination precision via function Mv::changePrecision. Basically, all interpolation processes use MV_PRECISION_INTERNAL (1/16-pel), all AMVP signaling related processes use MV_PRECISION_QUARTER (1/4-pel) as MVd signaling is at 1/4-pel. From local tests, this version generated identical results to the previous version in master (a86ab0).