[FEATURE] Categorizing models by supported modalities

[lthoang]

Description

We need a document for models' supported modalities for ease of model selection.

[darrylong]

I suggest that we could have another table, similar to in PR #624 and #632 that will be on readthedocs as well (below the original table).

This table could show models with different modalities supported (similar to the PyTorch/TensorFlow columns).

In order to obtain the modalities supported, we could run a script to go through the model files to look for the following keywords (e.g. 'sentiment', 'user_image'), as detailed in the BaseMethod file below.

cornac/cornac/eval_methods/base_method.py

Lines 281 to 290 in ef9956e

	self.user_feature = kwargs.get("user_feature", None)
	self.user_text = kwargs.get("user_text", None)
	self.user_image = kwargs.get("user_image", None)
	self.user_graph = kwargs.get("user_graph", None)
	self.item_feature = kwargs.get("item_feature", None)
	self.item_text = kwargs.get("item_text", None)
	self.item_image = kwargs.get("item_image", None)
	self.item_graph = kwargs.get("item_graph", None)
	self.sentiment = kwargs.get("sentiment", None)
	self.review_text = kwargs.get("review_text", None)

[tqtg]

We've somehow tried to achieve this via model type in the main README, though it's not complete. We can consider adding columns to the model table in the docs, but be careful of information overload.