Feat: add support for CSV input data loading from file or inline

[themisvaltinos]

This is WIP. Feedback is welcome.

• Adds support for CSV external file loading and inline CSV rows, as well as yaml from external file.
• Still to add tests and Docs to be updated accordingly.

EDIT (George):

This addresses the last item in #1637!

[CLAassistant]

All committers have signed the CLA.

[georgesittas]

Cool, looks good. Some general comments to keep in mind w.r.t. code style & conventions. The direction here is good, let's add some tests.

[georgesittas]

@Themiscodes let's make sure to test NULL behavior in CSVs as well. In YAML, when you leave out a column from a row, set its value to null, or leaving its value unspecified (i.e. col: ), this results in a NULL value when converting it. We should think about how we want to treat empty CSV values.

The problem is that ,, is ambiguous - it can mean both "empty string" and "null value". Perhaps we want to go with the latter here and if users ever need empty strings, they'd have to use a different format or something.

[themisvaltinos]

@Themiscodes let's make sure to test NULL behavior in CSVs as well. In YAML, when you leave out a column from a row, set its value to null, or leaving its value unspecified (i.e. col: ), this results in a NULL value when converting it. We should think about how we want to treat empty CSV values.

The problem is that ,, is ambiguous - it can mean both "empty string" and "null value". Perhaps we want to go with the latter here and if users ever need empty strings, they'd have to use a different format or something.

I agree that the latter option seems better. Currently empty columns are handled similar to the yaml and are set to nan. If the user wants an empty string it seems more reasonable that they should declare it explicitly.

[georgesittas]

@Themiscodes thanks for addressing comments, did another pass on this. Comments are mostly about refactoring the unit tests.

Let's also test 1) specifying paths (both csv, yaml) - look into tmp_path etc in other tests 2) doing format: yaml and then supplying the rows like we usually do (rows: ...), 3) supplying a path without a format (expected is to load yaml) 4) supplying a format that is not supported

[themisvaltinos]

@georgesittas addressed your comments

1. specifying paths (both csv, yaml) - look into tmp_path etc in other tests 3) supplying a path without a format (expected is to load yaml)
   in: test_format_path

2. doing format: yaml and then supplying the rows like we usually do (rows: ...),
   in: test_format_inline

3. supplying a format that is not supported
   in: test_unsupported_format_failure

Also removed unnecessary tests as per your suggestions and parametrized dictionaries where possible.

[georgesittas]

Suggested a couple of minor refactors, but otherwise LGTM. Thanks for addressing all comments @Themiscodes! 🙌