Round offsets to integral values

[connormanning]

COPC validation (link) was identifying data which was unpacking to be outside of its bounds due to the offset not being evenly divisible by the scale factor. For example, the following value:

Scale: 0.01
Offset: [215957.75236873244, 2034592.0870288194, 29.55000000000001]
Bounds (of node): [
    211645.27529721166,
    2033409.4779929873,
    -65.85000000000036,
    211739.02490434938,
    2033503.2276001251,
    27.8996071377328
]
Point: [211708.50236873244, 2033409.4770288193, -21.499999999999993]

Here, the Y-value of this point is less than the minimum-Y value of the bounds of its containing node. It may be true that it's not strictly incorrect, but for practical purposes pinning to a neater number so the associated floating point operations with bounds bisecting etc. will work better.

[connormanning]

Probably the comment here is still inaccurate, this PR does not attempt to preserve things like scale=0.1 offset=50.042 for example, where the source data would have raw values which are not pinned to the 0.1 grid.