Correction to frame standardization

[sedonaprice]

Fixes #703.

Changes the method SphericalSkyRegion._standardize_frame(), to return an instance, not a frame, and to first check if the frame is already an instance of BaseCoordinateFrame.

Test case for that issue now works as expected:

from astropy import units as u
from regions import RangeSphericalSkyRegion

reg1 = RangeSphericalSkyRegion(
    frame="icrs",
    latitude_range=[0 * u.deg, 1 * u.deg],
)
reg2 = RangeSphericalSkyRegion(
    frame="icrs",
    latitude_range=[0.5 * u.deg, 1.5 * u.deg],
)
reg1 & reg2

returns

<CompoundSphericalSkyRegion(region1=Region: RangeSphericalSkyRegion
frame: icrs
longitude_range: None
latitude_range: [<Quantity 0. deg>, <Quantity 1. deg>], region2=Region: RangeSphericalSkyRegion
frame: icrs
longitude_range: None
latitude_range: [<Quantity 0.5 deg>, <Quantity 1.5 deg>], operator=<built-in function and_>)>

[larrybradley]

Thanks, @sedonaprice. The fix looks good. However, this PR needs a regression test for the bug being fixed.

(I also pushed a small fix to your branch).

[sedonaprice]

@larrybradley Regression tests added (after a compound spherical sky test refactor as an independent commit, enabling more clear demarcation of the regression test).

The regression test I added for confirming the frame is standardized as expected for all shapes needed a carveout for WholeSphericalSkyRegion. Another option would be to do a minor refactor to introduce a "frame" for a WholeSphericalSkyRegion class. This would have a couple of benefits: (a) consistent behavior with the other region "shapes" (for tests), (b) opportunity for sanity checks (can the whole sky be transformed to X frame?). Something to consider for a different PR.

[larrybradley]

LGTM! Thanks, @sedonaprice.