Investigate "Categorical" vs "category" usage in df.ww.select across EvalML

While integrating Ordinal usage into AutoML via the Ordinal Encoder, I've noticed a couple of places that aren't treating Ordinals as a categorical column because they are using the "categorical" string in woodwork selects, which will only select columns that have the `Categorical` logical type, ignoring any other columns that are categorical in nature but may have different logical types (`Ordinal`, for example).

Instead, we should be using the `'category'` string, which will select any column that contains the "category" standard semantic tag. (See woodwork guide on [types and tags](https://woodwork.alteryx.com/en/stable/guides/logical_types_and_semantic_tags.html#Semantic-Tags))

We should look for places in EvalML that are using the "categorical" string and determine if they should be "Category" instead. This will be important as we continue to increase Ordinal usage in EvalML.