feat(python, rust): added new Config formatting option set_tbl_column_data_type_inline, fixed reading of env vars, improved interaction between formatting options

[alexander-beedie]

Final piece (probably?) of Config overhaul/enhancement ;)

Updates:

• Fixed reading boolean environment vars from Rust (was previously just checking if a var was set using is_ok(), but was not actually checking its value, so you couldn't set to "0", you'd have to completely unset it - which doesn't play well with saving explicit config options, state, context-scope, etc). Now respects both "1" & "0" values for all Config boolean vars and handles them appropriately.

• Added a lot of before/after table formatting examples to the docstrings, showing the visual impact of the given option.

• Fixed various format interactions - for example: if you dropped column types you don't want the --- dtype separator to persist, so now automatically dropping that as well (inferring a few other similar cases - leads to more consistent/pleasant formatting when setting a few different options).

• Renamed set_tbl_hide_column_separator to set_tbl_hide_dtype_separator, as the initial name is misleading (the separator being hidden has nothing to do with columns - it's the one between the name and the dtype). Technically breaking, but I suspect the number of affect people is going to be zero, or very very close to zero, so... ;)

• Removed set_utf8_tables() as this is never actually used - polars table format defaults to utf8; the only reason to change it to ascii is because you can't output utf8 strings (primitive tty/etc), in which case you are very unlikely to change it back.

  If for some reason you find you do want to do this, can just use set_ascii_tables(False) instead, which will revert formatting back to utf8; no need for a redundant method. The other useful ascii formats (such as markdown) are accessible through a different option anyway - for example: set_tbl_formatting("ASCII_MARKDOWN").

• Removed set_tbl_change_column_data_type_position_format; the name isn't clear and it's also redundant because you can get the same effect through the better-named set_tbl_hide_dtype_separator.

• New formatting option set_tbl_column_data_type_inline(True), which inlines the dtype after the name; trades some horizontal space for some vertical space:

  # shape: (3, 2)        shape: (3, 2)
  # ┌─────┬───────┐      ┌───────────┬────────────┐
  # │ abc ┆ xyz   │      │ abc (f64) ┆ xyz (bool) │
  # │ --- ┆ ---   │      ╞═══════════╪════════════╡
  # │ f64 ┆ bool  │      │ 1.0       ┆ true       │
  # ╞═════╪═══════╡      ├╌╌╌╌╌╌╌╌╌╌╌┼╌╌╌╌╌╌╌╌╌╌╌╌┤
  # │ 1.0 ┆ true  │  >>  │ 2.5       ┆ false      │
  # ├╌╌╌╌╌┼╌╌╌╌╌╌╌┤      ├╌╌╌╌╌╌╌╌╌╌╌┼╌╌╌╌╌╌╌╌╌╌╌╌┤
  # │ 2.5 ┆ false │      │ 5.0       ┆ true       │
  # ├╌╌╌╌╌┼╌╌╌╌╌╌╌┤      └───────────┴────────────┘
  # │ 5.0 ┆ true  │
  # └─────┴───────┘

• Additional test coverage: validates the new formatting options, context-scoping, and the interaction of different formatting options applied at the same time.

Misc:

• The Config doctest for set_tbl_cols was leaking its environment env change into other tests; once I stopped that, the to_dummies doctest started failing as it was implicitly relying on the higher (non-default) number of columns that got set there; fixed that by slightly modifying the example.

[alexander-beedie]

The updated docstrings need a bit more work to pass tests :)
Will finish it up tomorrow.

Update: done.

[ritchie46]

Huge configuration improvement. Thanks!