Fourier class sample_prior support for non-Prior classes

[wd60622]

Having sample_prior method is not requirement of VariableFactory, only Prior so this needs to be changed:

pymc-marketing/pymc_marketing/mmm/fourier.py

Lines 460 to 478 in c989eeb

	def sample_prior(self, coords: dict | None = None, **kwargs) -> xr.Dataset:
	"""Sample the prior distributions.
	Parameters
	----------
	coords : dict, optional
	Coordinates for the prior distribution, by default None
	kwargs
	Additional keywords for sample_prior_predictive
	Returns
	-------
	xr.Dataset
	Prior distribution.
	"""
	coords = coords or {}
	coords[self.prefix] = self.nodes
	return self.prior.sample_prior(coords=coords, name=self.variable_name, **kwargs)

Something like in the media transformations will have to be done in order to support non-Prior classes:

pymc-marketing/pymc_marketing/mmm/components/base.py

Lines 339 to 361 in c989eeb

	def sample_prior(
	self, coords: dict | None = None, **sample_prior_predictive_kwargs
	) -> xr.Dataset:
	"""Sample the priors for the transformation.
	Parameters
	----------
	coords : dict, optional
	The coordinates for the associated with dims
	**sample_prior_predictive_kwargs
	Keyword arguments for the pm.sample_prior_predictive function.
	Returns
	-------
	xr.Dataset
	The dataset with the sampled priors.
	"""
	coords = coords or {}
	dims = tuple(coords.keys())
	with pm.Model(coords=coords):
	self._create_distributions(dims=dims)
	return pm.sample_prior_predictive(**sample_prior_predictive_kwargs).prior