Replace localization uncertainty properties with one error radius

[lpsinger]

Replace the localization properties uncertainty_shape, ra_uncertainty, dec_uncertainty, and position_angle with a single property called ra_dec_error.

The original localization uncertainty properties were ill-conceived in many ways:

1. The names ra_uncertainty and dec_uncertainty were misleading. For example, for ellipses, these fields actually represented the lengths of the semi-major and semi-minor axes, and not the uncertainty in ra or dec.

2. The names of these properties were inconsistent with other schema. Every other uncertainty property exept for these has a name ending in _error, not _uncertainty.

3. The rectangle shape was not useful. The most significant examples of quadrilateral localization regions are IPN error boxes. However, these are parallelograms, not rectangles, so the shape would have been under-specified.

We have observed that all current missions in GCN either report error circles or HEALPix maps. Given that observation, this patch replaces all of the ill-conceived localization properties with a single ra_dec_error field which is the radius in degrees of the error circle.

In the future, there might be a mission that requires ellipses. We won't deal with that now, but in that event we would have a few different backwards-compatible options. For example, we could:

• Treat the ra_dec_error field as the semi-major axis, and add optional fields ra_dec_error_b and ra_dec_error_pa for the semi-minor axis and the position angle, respectively; or

• Allow ra_dec_error to be either a single scalar, or an array of up to three numbers, representing the semi-major axis, semi-minor axis, and position angle, respectively.

Fixes #120.

[Vidushi-GitHub]

I would suggest to re-think, to my knowledge, the only circular localization was limitation of the old GCN. If we believe that ellipse would be convenience for some of the users eg, GW, FRB, We should come up with creative idea during implementation, such that both the communities are happy.

[jracusin]

I would prefer to keep the ellipse option in there so that we're planning ahead. I like @lpsinger's ra_dec_error second option with either a single scalar or an array of up to three numbers.

[lpsinger]

I would prefer to keep the ellipse option in there so that we're planning ahead. I like @lpsinger's ra_dec_error second option with either a single scalar or an array of up to three numbers.

Alright. @Vidushi-GitHub, please update your PR accordingly.