Consolidate dataframes DSL into select

[ericpan64]

Closes #9

Background

Dataframe DSL didn't really cut-down on boilerplate code and wasn't that different from polars/pandas APIs as-is (which have stronger typing guarantees). To differentiate this a bit, moved a lot of functionality into select and a string-based syntax which also matches

Design (high-level)

• Added a default syntax for specifying tables: A, B, ... , Z. In-order based on the dataframes passed in the new param others in the select function
  • Skip named variables for now -- don't think people care nor is it helpful (take stronger opinion as framework)
• Added a from clause in select syntax as follows:
  • Moved join into from A <- B on [colname], <- for left join, <> for inner join
  • Moved group_by into from A => groupby[colname | agg_fn] which uses all() by default

Other notes

• Removes +> syntax which didn't seem that useful (though good to think through). This really simplifies the code which is worth it
• A lot of this is hacky and will be re-implemented with a better string parser. Will want to think-through the design of that more thorughly
  • This will allow for generic expressions, subqueries, etc.
  • Also a good time to add *-based expressions (slicing, "all columns except", regex match, etc.)