Overhaul SE

[jkanche]

This branch is almost a rewrite of the entire package.

• Implements functional paradigm to access and set attributes.
• Can initialize SE/RSE without assays, also initialize an empty SE/RSE
• row_data and column_data are expected to be BiocFrame objects and will be converted if a pandas DataFrame is provided. This will allows us to reduce complexity and provide consistent downstream operations. If these are not provided, an empty BiocFrame is set as default.
• On RSE, if row_ranges is not provided, an empty GenomicRangesList is set as default.
• SE/RSE now contain row_names and column_names that are separate from row_data's and column_data's row names, also helps in simplifying subset operations.
• Printing SE/RSE objects now looks almost similar to R/Bioc's printing of these objects.
• Support combine operations, both strict and a flexible combine option when rows or columns do not exactly match.
• Streamlines subset operation for both SE & RSE.
• Reduce dependency on a number of external packages.
• Update docstrings, tests and docs.

[jkanche]

@LTLA do the combine functions make sense?

[jkanche]

FYI: @keviny2 this is finally done. checkout combine_* and relaxed_combine_* operations. you may want to check these out for your package.