add method to compute average treatment effects

[apoorvalal]

Thanks for the nice work on this package; it addresses a lot of pain-points I have with econML. An ATE estimator method was missing (possibly because that might not be the most obvious intended use case for the package, but the rest of the API / package ergonomics are nice enough that I decided to use it for that as well), so I added it (only to drlearner, since the variance-of-pseudooutcome construction for standard error is justified in that case but not for T-learner etc).

I've also added a notebook in the docs/examples directory that verifies that estimates line up with what one gets from econML and doubleML on a simulated dataset.

Checklist

• Added a CHANGELOG.rst entry

[apoorvalal]

build failures likely due to use of external packages in the new notebook. Should I add them as dev dependencies (or I can install them inside the notebook; unsure if that conforms to your style guide).

Edit: done

[kklein]

@apoorvalal Thanks a lot for your kind words and effort! :)

I'll get to your PR asap - hopefully on the weekend!

[kklein]

Looking great - just some minor comments. :)

[kklein]

A simple pixi update should suffice to get rid of the merge conflict. :)

[apoorvalal]

addressed most of your (great) suggestions in this commit. I think the conflicts are likely straightforward to resolve (amended a the drlearner test file, added the feature desc to changelog, and the usual pixi lock conflict

[apoorvalal]

Ready to merge pending (opaque) pixi trouble (edit: and mypy decided to get mad at a different commit]
I added you as a collaborator on my PR so you can directly commit to my ate branch rather than to have to send me PRs.

[kklein]

LGTM! Thanks a lot for your contribution @apoorvalal ! :)

I'll create a new release for version 0.9.0 soon.