Fixes #279 and #100 plotting updates

[bnubald]

Resolves #279 and #100

• Uses Cartopy to enable north-facing plots.
• Enable defining region for forecast output based on lat/lon bounds and not just pixel bounds.
• Enable gridlines optionally.
• Enable coastlines with mp4 animation output (Fixing xarray_to_video and coastlines #100).

There is still work to be done on using the lat/lon which will likely need some refactoring of the Masks class. When process_regions is called and Masks is provided to it here:

icenet/icenet/plotting/forecast.py

Lines 1486 to 1488 in f128536

	if args.region:
	seas, fc, obs, masks = process_regions(args.region,
	[seas, fc, obs, masks])

It defines a pixel based slice which is set by getitem in the Masks class. (This is the case for all times process_regions is called in forecast.py above)

icenet/icenet/data/sic/mask.py

Lines 305 to 315 in f128536

	def __getitem__(self, item):
	"""Sets slice of region wanted for masking, and allows method chaining.
	This might be a semantically dodgy thing to do, but it works for the mo
	Args:
	item: Index/slice to extract.
	"""
	logging.info("Mask region set to: {}".format(item))
	self._region = item
	return self

But, the Masks class needs to account for when the bounds are defined by lat/lon. This is needed when the Mask is used to weight the metric, for example:

icenet/icenet/plotting/forecast.py

Lines 88 to 100 in f128536

	# obtain mask
	agcm = masks.get_active_cell_da(obs_da)
	# binary for observed (i.e. truth)
	binary_obs_da = obs_da > threshold
	# binary for forecast
	binary_fc_da = fc_da > threshold
	# compute binary accuracy metric
	binary_fc_da = (binary_fc_da == binary_obs_da). \
	astype(np.float16).weighted(agcm)
	binacc_fc = (binary_fc_da.mean(dim=['yc', 'xc']) * 100)