feat(python): add datetime/duration dtype selector groups covering the different timeunits

[alexander-beedie]

ref: #6422, #5300

Extends the available selector groups added in #6295, with DATETIME_DTYPES and DURATION_DTYPES; these each cover the ms, us, and ns time units so you can more ergonomically select any datetime/duration col without having to specify them explicitly (note: ideally we still need a wildcard for timezones - "to be continued...")

[ritchie46]

I think this is valuable. 👍

As we are getting more default globals, could we look into adding a test that keeps checking our import times?

[alexander-beedie]

As we are getting more default globals, could we look into adding a test that keeps checking our import times?

We should probably go all the way and start a benchmarking/timing suite (that would include import time) that covers all of the basics (eg: do some large aggregations, joins, unions, iteration, etc) and grow it over time 🤔

Could be something that we run before each release to catch performance regressions.

If there aren't plans for this yet I can take a quick survey over other projects to see how they do the same (V8 probably being the most famous example ;)

[ritchie46]

We already run a benchmark on every PR. That compiles python polars in release mode, so we can extend the benchmarks that run afterwards.

[ritchie46]

If there aren't plans for this yet I can take a quick survey over other projects to see how they do the same (V8 probably being the most famous example ;)

Very welcome!