Rewrite using Polars

[blaylockbk]

I'm inclined to re-write this using Polars. I love Polars!

• [x] Load data into Polars DataFrames.
• [ ] Timeseries data will return be in one dataframe rather than a list of dataframes. (Use categorical dtype for some columns like STID, TIMEZONE, etc.)

[blaylockbk]

Note to self:

Saving a JSON copy of the returned data that for 18 stations, all variables for 1 month is ~25 MB on disk. Organizing the data into a Polars DataFrame and saved to Parquet is 131KB.

[blaylockbk]

I'm making great progress on this. Need a todo list

Code

• [x] Data will be provided in long format by default. Add optional argument to pivot the data.
• [ ] add optional argument to return data as Pandas data frame (for those users who prefer pandas, but I'm telling you that I am fully on the polars bandwagon; I don't use pandas anymore)
• [ ] add optional argument with_latency
• [ ] basic plotting for summary; seaborn will be an optional dependency

Docs

• [x] rewrite docs with new examples
• [ ] examples of doing rolling/resample windows
• [ ] examples of pivot long to wide format
• [ ] examples of plotting with seaborn
• [ ] show users how to save to parquet (and the benefits of doing so)
• [ ] rewrite readme

GitHub

• [x] explain big overhaul to users. The entire package is a breaking change; it's practically a new package. Reasons: improve maintainability, I wanted to learn polars, I am learning class Inheritance, long format data frame makes more sense to me.