Del _FillValue from coordinates at cf_encoder

[BorjaEst]

• Closes to_netcdf -> _fill_value without NaN #2037
• Tests added
• User visible changes (including notable bug fixes) are documented in whats-new.rst
• New functions/methods are listed in api.rst

[BorjaEst]

Probably there is a better way to solve this.
It also gets in conflict with backend tests (which conflict CF conventions as described at the comments).

[dcherian]

Thanks for trying to fix this longstanding issue @BorjaEst.

I've added a couple of comments. Please let us know if you need help finish it.

I imagine some of the tests will need to be updated to remove the _FillValue attribute

[dcherian]

1. Xarray does allow dimension coordinates with NaNs. So we'll need to check for the presence of nans using np.any(duck_array_ops.isnull(new_vars[var])). If there are no NaNs then we avoid the _FillValue.
2. Can we move this logic to

   xarray/xarray/coding/variables.py

   Lines 144 to 148 in 2c68a8b

   	class CFMaskCoder(VariableCoder):
   	"""Mask or unmask fill values according to CF conventions."""
   	def encode(self, variable, name=None):
   	dims, data, attrs, encoding = unpack_for_encoding(variable)

3. IIRC we always add _FillValue so the second half of the condition can be removed.

[BorjaEst]

From CF terminology:

coordinate variable
We use this term precisely as it is defined in the NUG section on coordinate variables. It is a one-dimensional variable with the same name as its dimension [e.g., time(time) ], and it is defined as a numeric data type with values that are ordered monotonically. Missing values are not allowed in coordinate variables.

I suppose the best is raise an Excepton ValueError: Missing values are not allowed in coordinate variables on the logic you indicated if is a coordinate variable and there are null_values.

[BorjaEst]

2. Can we move this logic to xarray/xarray/coding/variables.py

I tired, but in the variable context, it is not possible to access to bound variables (which are kind of coordinate).
I also tried to modify here:

xarray/xarray/conventions.py

Lines 125 to 133 in 33cdabd

	def maybe_default_fill_value(var):
	# make NaN the fill value for float types:
	if (
	"_FillValue" not in var.attrs
	and "_FillValue" not in var.encoding
	and np.issubdtype(var.dtype, np.floating)
	):
	var.attrs["_FillValue"] = var.dtype.type(np.nan)
	return var

But I think that overcomplicates the logic later if we have to search for the bounds.

Any ideas?