Thanks for this PR @savente93! A couple of comments to discuss

• You deviated from the approach of other model properties by not implementing a dynamic (not sure this is the right terminology) public tables property and initiating the internal _tables property with None. Using a dynamic tables property allows to trigger reading upon first call (i.e. when the _tables property is still None; this should also be changed). See e.g. Model.maps property.
• You also implemented a different get methods and an iter methods which we don't have for other properties. I think we can keep those but these should point to the tables instead of _tables property to trigger the reading.
• I'm also not sure if we need to support different drivers in the read and write methods here. Most models will probably use a custom format and hence overwrite the read and write methods. Thus writing to different table formats like excel and/or parquet will likely not be used. I suggest to limit it to a single csv format similar to the other properties to keep thinks simple on our side.

I've made some final changes to

1. make the method between properties more consistent, i.e., order of arguments in set_tables; usage of single set_tables method rather than singular and plural; and fn argument in read and write.
2. add tables to default read and write components all (sub)model classes.
3. added tables to append test