To avoid precision problem of float, such as
(int)(960/1.2f)=799
(int)(960/1.2f+0.5f)=800
(int)(540/1.2f)=449
(int)(540/1.2f+0.5f)=450
Bug: 204481252
Test: DisplayContentTests#testCreateTestDisplayContentFromDimensions
Change-Id: I8f78d9bd14b7fb9d12a20a662ad4ec04f82ce7f2