Numerblox

What's Changed
* Churn by CarloLepelaars in https://github.com/crowdcent/numerblox/pull/170


**Full Changelog**: https://github.com/crowdcent/numerblox/compare/v1.3.3...v1.3.4

What's Changed
* Compute exposures with lstsq (~2x speedup) by CarloLepelaars in https://github.com/crowdcent/numerblox/pull/169. Exposure speedup tricky thanks to **by256** in forum post: https://forum.numer.ai/t/2x-feature-neutralization-speed-up/4473.
* Relax versions for `numerai-tools`, `numerai-era-data`, `scikit-learn`, `numerapi` and `numpy`.

**Full Changelog**: https://github.com/crowdcent/numerblox/compare/v1.3.2...v1.3.3

What's Changed
* Security updates based on Github Dependabot
* Add support for v5 data by CarloLepelaars in https://github.com/crowdcent/numerblox/pull/168

**Full Changelog**: https://github.com/crowdcent/numerblox/compare/v1.3.1...v1.3.2

What's Changed
* Feature/no era routing by CarloLepelaars in https://github.com/crowdcent/numerblox/pull/165


**Full Changelog**: https://github.com/crowdcent/numerblox/compare/v1.3.0...v1.3.1

What's Changed
* Update/sklearn14 by CarloLepelaars in https://github.com/crowdcent/numerblox/pull/163
* Feature/metadata routing default by CarloLepelaars in https://github.com/crowdcent/numerblox/pull/164


**Full Changelog**: https://github.com/crowdcent/numerblox/compare/v1.2.2...v1.3.0

- Support for [scikit-learn's metadata routing](https://scikit-learn.org/stable/metadata_routing.html).
- Updated tests and documentation for metadata routing.
- Support for [Polars](https://pola.rs/) output in Numerblox processors.
- Metadata routing is initialized so does not have to be set explicitly. Can still be switched to False or None for explicit cases where additional arguments are not needed. For example, in live inference where era_series is a constant series (1 era) so it would be redundant to pass.

NOTE: Numerblox uses metadata routing, which was introduced in `scikit-learn` 1.4 and is still in an experimental phase. Not all `scikit-learn` meta-estimators work with metadata routing yet. Check [this overview](https://scikit-learn.org/stable/metadata_routing.html#metadata-routing-support-status) for the support status on scikit-learn meta estimators that support metadata routing.

For most cases with Numerblox processors you don't have to set any routing requests manually. It is already handled within the processors. For more information about metadata routing:
https://scikit-learn.org/stable/metadata_routing.html

**_WARNING:_** `eras` argument is now `era_series` and `tickers` argument is now called `ticker_series`. This is changed to clarify that the input argument should be a `pd.Series`.

**Full Changelog**: https://github.com/crowdcent/numerblox/compare/v1.2.1...v1.2.2