Fix rounding error when constructing Rect from FloatRect

Current rounding logic does not handle negative values
correctly. FloatRect(-1080,...) is converted to Rect(-1079,...)
Fix this by using the standard rounding function.

Flag: EXEMPT build error (b/350967139)
Bug: 310950423
Test: presubmit
Change-Id: Ib8c5c393040149225e5b2bc858c21894e306a456