Use Literal for statistics.quantiles()

[Akuli]

Fixes #7462

[JelleZijlstra]

Looks good. I don't think #7258 needs to block this.

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

core (https://github.com/home-assistant/core)
+ homeassistant/components/statistics/sensor.py:646: error: Argument "method" to "quantiles" has incompatible type "str"; expected "Union[Literal['inclusive'], Literal['exclusive']]"  [arg-type]

[hauntsaninja]

Given that there are false positives in practice, "prefer false negatives over false positives" would imply we shouldn't do this without a fallback overload right?

[AlexWaygood]

Given that there are false positives in practice, "prefer false negatives over false positives" would imply we shouldn't do this without a fallback overload right?

How would we do a fallback overload in this case?

[AlexWaygood]

I think this is a good change to make. The one false positive is very easily fixed. #7258 is a temporary moratorium on literal constants, not on literal argument types.

[hauntsaninja]

It'd be a bad fallback in that we'd have zero additional type safety but would still mean we get better IDE hover docs / completion docs :-)
I don't feel too strongly, so feel free to merge, but there is still a similar consideration on "soft literal"s here. The ideal type checker behaviour for me here is "if we know the value of the string, make sure it matches the literals. if we don't know, allow it", and we don't have a way of spelling that currently.

[AlexWaygood]

Shoot, I screwed up the commit description a bit there. Apologies :)

[Akuli]

It happens :) It's not super critical with stdlib stubs anyway. With third-party stubs, this matters more because commit messages turn into an auto-generated changelog.