Sky grid fixes

[pannarale]

Regarding #4485 , let's

• rely on already existing functions for angular/cartesian conversions, if possible
• check conventions (coordinates especially)
• fix a small boundary effect issue with the current circular grid
• improve beyond circular uniform grid, ideally with the ideas in https://www.arxiv.org/abs/1208.2340 and https://arxiv.org/abs/1307.4158 or with the approximate (but less approximate than what is currently implemented) https://arxiv.org/abs/1509.03426
• PEP8 style. This can largely be done automatically, see https://github.com/gwastro/pycbc/wiki/How-to-satisfy-CodeClimate.

[titodalcanton]

PEP8 should have (mostly) been addressed in #4988.

[pannarale]

• At line 90, sky_grid.py does detectors = hf.attrs['detectors'], but pycbc_make_sky_grid fills ins hf['instruments'] = [d for d in args.instruments] at line 187. So if you generate a sky grid with pycbc_make_sky_grid and pass it to pycbc_multi_inspiral, it will fail because sky_grid.py expects 'detectors'. I leave it to @titodalcanton to decide which of the two ends to modify.

P.S. It would probably be better to have pycbc_make_sky_grid use the sky_grid.py module when writing the grid file to keep things centralised. Poking around, changes may have an impact on pycbc_multi_inspiral (which reads the sky grid and computes time delays and antenna patterns), and on pycbc/workflow/grb_utils.py (when it makes the sky_grid node) and pycbc/workflow/jobsetup.py (when it makes the multi_inspiral nodes).

[titodalcanton]

@pannarale indeed, thanks for catching that. And a similar problem exists with the ref_gps_time / trigger_time attributes.

[titodalcanton]

#5000 ticks the second and last checkboxes.

[pannarale]

For reference, here is a run on multiple sky points.