Refactor ingestion, save and load functions

After the merge of #2391 we will have the following functions:

• load_edges
• load_nodes
• load_edge_metadata
• load_node_metadata

These will handle any dataframe inputs including (tested):

• Pandas dataframes
• FireDucks(.pandas) dataframes
• Polars dataframes
• Arrow tables
• DuckDB (eg. DuckDBPyRelation obtained from running an SQL query)

In addition to this, we should modify the function to internally check if the user has provided a dataframe or a filepath as an argument and transparantly handle loading from this if possible. Through this we can remove the following functions as they would be rolled into the above:

• load_edge_props_from_pandas
• load_edge_props_from_parquet
• load_edges_from_pandas
• load_edges_from_parquet
• load_node_props_from_pandas
• load_node_props_from_parquet
• load_nodes_from_pandas
• load_nodes_from_parquet

To reduce confusion we will change the following:

• load_from_file -> load_graph
• save_to_file -> save_graph
• save_to_zip -> flag on save_graph (to be discussed)

Post v4 merge the following functions will be removed:

• from_parquet/to_parquet -- as we are removing the proto format and this will become the default format for save/load_graph
• load_cached/write_updates/cache -- we have the WAL now