[FEA] Support RoundCeil and RoundFloor when scale is zero

[nartal1]

Is your feature request related to a problem? Please describe.
We have currently disabled running RoundCeil and RoundFloor expressions on GPU and fall back to CPU for all inputs. We can enable it for scale=0 as it would call GpuCeil/GpuFloor internally. There is a bug where it produces incorrect results when the input type is DecimalType(precision = 18, scale = 0) .
We should enable these integration tests to verify:

1. test_floor_scale_zero
2. test_ceil_scale_zero