Change dataset option in GFD during python integration

[Mstrutov]

The reason we have a separate option Table for CSV tables is because the table can be specified for the algorithm not only as path, but also as pandas dataframe. So what we need to think about here is how do we want to specify input graphs from python, if using path to the file, then storing only path is ok.
Since in this PR we're not touching python bindings at all, there is little reason to create a separate option, and we can just store paths for simplicity. So IMO the separate option, if required, should be added in the future along with python bindings.

Originally posted by @polyntsov in #154 (comment)