Use new data model differentiating single line loop blocks from simple blocks

[alisterburt]

This library uses a dict of dataframes as the in memory data model for a STAR file. There is an ambiguity when dealing with single row dataframes, are they simple blocks or loop blocks?

This PR starts work on changing the underlying data model to:

• simple blocks: dict[str, str | int | float]
• loop blocks: pd.DataFrame (as before)

Thus removing the ambiguity. This will break the API and be released as 0.5.0 when ready

Parser has been heavily reworked and simplified, all parsing tests are passing - haven't touched the writer or the high level functions yet but the bulk of the work is done

[alisterburt]

functional interface now at feature parity for both reading and writing

todo:

• add explicit tests for writing from dict data blocks
• make sure typing backports work as expected