Add a `tolerance` parameter to `MaskedGrid`

[jjrob]

Currently, MGET's MaskedGrid requires the extents of the masked grid and the masks to have exactly the same extents and CenterCoords. Occasionally, when using different software to produce the grid and the masks, or even when using ArcGIS to produce both, the extents differ by a small value like 1e-9, presumably due to the imprecision of IEEE 754 floating point numbers causing a difference based on the specific way calculations are done (e.g. different orders in arithmetic operations). We should add a tolerance parameter that allows the validations of the grid and masks to succeed even when they are "off" slightly because of this problem.