Move pandas dataframe handling to external convert_dataframe module

[CalCraven]

This PR looks to improve the handling for converting a topology to a dataframe. This currently lives as a method for topology. It is now being moved to a convert_dataframe.py module. A few different formats are available which give some nice default ways to view a topology. Notably, we have the formats:
-publication which gives all the parameter values you would want to have in a table for publication. This also removes duplicates so each parameter is only listed once.
-default some default values which are nice to have
-remove_duplicates which allows you to get a smaller dataframe with duplicate rows removed.
-specific_columns Allows the user to specify what they want in the dataframe.

There is also an added function that allows you to generate dataframes that cover the parameters for a set of topologies.

Finally, there will be some function that prints the dataframes with the rdkit mols which are labeled to match the dataframes.

TODO Checklist:

• Error checking on arguments
• Replace topology.py dataframe methods/tests
• Doc strings
• Handle units
• Handle parameters that return lists
• Handle parameters that return dictionaries
• Return unique elements if style is publication
• Handle parameter "all" better
• Function to concatenate multiple topogies into one output
• Remove replicate rows flag -> similar to the publication style, but without the atom_indices section added
• Function to create a topology with all data from dataframe matching rdkit mol image
  • Could just also make this a format unique_types

[chrisjonesBSU]

I left some comments after scrolling through the changes. I still need to test this out in a notebook though. I'll be sure to leave another review and/or comment after I do.

Thanks for doing this Cal, this will be a very useful feature!

[chrisjonesBSU]

Looks like there is an extra tab here

[chrisjonesBSU]

I'm kind of confused about this line, and I don't see anything in the notes about it. Is this saying that remove_duplicates specifically requires that you don't use all

[chrisjonesBSU]

Small typo here, I think this should say format

[chrisjonesBSU]

Also, we should add remove_duplicates as an option here correct? I think this error message could be a little more clear. Something like "Available options for format are 'default', 'specific_columns', 'publication', or 'remove_duplicates'."

Also, the f string isn't doing anything here, do we want to print out what they passed to format?

[chrisjonesBSU]

We could take this out.

[chrisjonesBSU]

Is this function called or used anywhere? Looks like this is the old method from Topology, I'm guessing we can remove it here?

[chrisjonesBSU]

We could remove this one as well

[chrisjonesBSU]

Do you have an example of how to see and access the units when using handle_unyts="in_headers"? It doesn't seem to be placing the units in the column names from what I can tell. It does work as expected when using with_data. Also, its not clear what the expected behavior of all_floats is from the doc strings. Does that mean not to include units at all? Maybe this could be more clear by using a different option name (e.g. no_units) or adding an extra sentence to the doc strings.

[chrisjonesBSU]

A couple more thoughts:

1. Do we want parameter to be passable as a list of parameters? Right now, it is either pick one, or get all of them. Would we prefer something like this to also work?

top_df_dict = to_dataframeDict(
    topology=top,
    parameter=["bonds", "angles"],
    handle_unyts="with_data"
)

We could handle this similarly to what we do with ignore_params in apply

2. This is a matter of preference and style, but these changes use a lot of camel case in variable names and the main function name (to_dataframeDict), and I think most of the gmso code base is snake case. Should we try to be consistent here as well?