Mocpy

Added

* Change API for ST-MOC: query_by_time, query_by_space

Added

* Space-Time coverages, classmethod from creating them from (time, ra, dec) tuples
* Query a Space-Time coverages with time frames and spatial coverages

Added

* Rust backend
* Add tests for windows py27 and py3
* from_polygon relies on cdshealpix. spherical_geometry dependency removed!
* change astropy-healpix dependency from mandatory to optional. astropy-healpix is now used in only a few deprecated methods
(such as the old plot method from `mocpy.MOC` and `get_boundaries` which will soon make use of cdshealpix too).

Changed

* API CHANGE!: the ``inside`` parameter of from_polygon and from_polygon_skycoord has been removed !
The inside of the polygon is deduced from the order of the sky coordinates passed.

Changed

* Change from_cells to from_healpix_cells. Its API does change too. It now takes the three ipix, depth and flags numpy arrays separatly instead as a numpy structure. This has the advantage of direcltly passing the arrays returned by `cdshealpix`. Creating a numpy structured array from simple numpy column arrays needed to copy the data from the arrays to the structural array.
* Add cdshealpix as a dep. Will remove astropy-healpix dep. When cdshealpix will be merged into astropy-healpix then the astropy-healpix dep will be restored.

Added

* Serialize to str. Call moc.serialize(format="str")
* Load a MOC from a str (see section 2.3.2 of [MOC IVOA paper](http://ivoa.net/documents/MOC/20190215/WD-MOC-1.1-20190215.pdf)).
* Fix vizualization bug when plotting all the allsky MOC. Biggest cells to plot are limited to depth 2. Cells of depth 0 and 1 are
subdivided into cells of depth 2 for the visualization purpose.

Changed

* Add of a `overwrite` optional keyword in the write method. Before 0.5.6 the default behaviour was to
always overwrite already existing files. Now it does not overwrite by default. To do that, you have to
set the `overwrite` keyword.

Added

* Plotting a moc with matplotlib axes is faster (concers **fill** and **border** methods). The level of detail of the plotted MOC is function of the FoV. The MOC is degraded to the minimum depth so that a cell of this depth can be contained in 1px at the center of the projection. For small FoVs, we only plot the part of MOC contained in the view (thanks to the speed of logical operation between MOCs).
* The [docs](https://mocpy.readthedocs.io/en/latest/) features more examples on how to plot a MOC, perform logical operations between MOCs, etc...
* The doc of the API has been reviewed and features some test codes that can be run with the sphinx command `make doctest`.

Removed

* The use of **multiprocessing** in the `fill` method.