Longitude normalization

[zmoon]

The longitude wrapping/normalization can introduce small floating point errors into longitude coordinates. For example:

>>> (120.711744 + 180) % 360 - 180
120.71174400000001

In some situations (no renames), _dataset_to_monet was doing this in place. That should now not happen. Additionally, currently I have it set to compute if lon normalization is necessary (min/max) and skip it otherwise, which seems advantageous.

In [35]: a = np.random.rand(int(1e7))

In [36]: %timeit (a.min(), a.max())
10.8 ms ± 60.7 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

In [37]: %timeit (a + 180) % 360 - 180
296 ms ± 3.08 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

cc: @rschwant @bbakernoaa