feat(python): Improved type-inference for read_excel and read_ods, use calamine engine for read_ods

[alexander-beedie]

Closes #15596.
Closes #15748.
Closes #15763.
Closes #11274.

Taking care of a number of paper-cut issues here; four for the price of one... ;)

Updates

• Exposes "infer_schema_length" parameter at the top-level of both read_excel and read_ods, automatically integrating it with engine-specific "read_options" (where available; otherwise it's passed down to DataFrame init).
• Improves initial parse quality of "calamine" (fastexcel) when type-inference is provided via "schema_overrides", translating the Polars dtypes to the dtype name expected by fastexcel's load_sheet_by_name method.
• Improves docstring/clarity for read_options, making it a lot clearer which method the options will be passed to.
• Updates the default engine for read_ods to "calamine"; this is not a breaking change as the "ezodf" engine was not configurable in any way, and read_ods has never had an "engine" parameter. Switching out internally for "calamine" removes a lot of custom parsing code, massively improves performance, allows for additional parsing options, and gets rid of an optional dependency on a package that hasn't had an update in about 10(!) years.
• Marks the "pyxlsb" engine as deprecated, with a suggestion to move to "calamine"; this will allow for further cleanups later. As with "ezodf" this engine has no configurability and is 10x slower than the "calamine" (fastexcel) alternative.
• Some minor drive-by improvements to the "source" parameter docstring (across all relevant methods).

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 80.40%. Comparing base (0c37ead) to head (64a1ab9).
Report is 3 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main   #15808   +/-   ##
=======================================
  Coverage   80.39%   80.40%           
=======================================
  Files        1264     1264           
  Lines      165421   165423    +2     
=======================================
+ Hits       132994   133008   +14     
+ Misses      32427    32415   -12     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ritchie46]

Cool stuff! Can we switch to calamine for excel as well?

[alexander-beedie]

Cool stuff! Can we switch to calamine for excel as well?

Definitely, and I plan to do so; fastexcel has had a number of good updates since we initially integrated it 😎

I think that it needs a bit more prep work than the two (more niche) engines I'm switching over here though, due to "xlsx2csv" being the default, and having a lot of customisation available via the read_csv params you can set via "read_options". I'm going to expose a few more options at the top-level to make the common-cases more accessible to all engines (selecting specific columns to read, etc), and then once that's done will make a breaking change to have "calamine" be the new default engine (while still respecting explicit opt-in to the other engines, of course) ✌️

[ritchie46]

All right. That sounds good. 👌😁