Hide internal methods for df management from the API and make them stateless

[pm-blanco]

Solves #117

Changed

• Methods that interact directly with the pyMBE dataframe are now private and stored in a dedicated module in storage/df_management. These methods also have been refactored to be stateless methods, i.e. making it impossible for them to change behavior during the pyMBE object lifetime or for the user to change the pyMBE dataframe unless explicitely calling them. This includes the methods: add_bond_in_df, add_value_to_df, assign_molecule_id, check_if_df_cell_has_a_value, check_if_name_is_defined_in_df, check_if_multiple_pmb_types_for_name, clean_df_row, clean_ids_in_df_row, copy_df_entry, create_variable_with_units, convert_columns_to_original_format, convert_str_to_bond_object, delete_entries_in_df, find_bond_key, setup_df.
• define_particle_entry_in_df is now a private method in pyMBE, as it is a convenience method for internal use.
• The custom NumpyEncoder is now a private class in the private module storage/df_management because it is only internally used in pyMBE for serialization/deserialization.

@jngrad this PR implies changes in many lines since I moved a lot of functions to the new module dedicated to the management of the pyMBE dataframe, but otherwise I have not changed the behaviour of pyMBE.

[jngrad]

Thanks!