Lifelines

New features
- adding `weights` to log rank functions

Bug fixes
- Fix using formulas with `CoxPHFitter.score`

Error in v0.26.1 deployment

API Changes
- `t_0` in `logrank_test` now will not remove data, but will instead censor all subjects that experience the event afterwards.
- update `status` column in `lifelines.datasets.load_lung` to be more standard coding: 0 is censored, 1 is event.

Bug fixes
- Fix using formulas with `AalenAdditiveFitter.predict_cumulative_hazard`
- Fix using formulas with `CoxPHFitter.score`

New features
- `.BIC_` is now present on fitted models.
- `CoxPHFitter` with spline baseline can accept pre-computed knot locations.
- Left censoring fitting in KaplanMeierFitter is now "expected". That is, `predict` _always_ predicts the survival function (as does every other model), `confidence_interval_` is _always_ the CI for the survival function (as does every other model), and so on. In summary: the API for estimates doesn't change depending on what your censoring your dataset is.

Bug fixes
- Fixed an annoying bug where at_risk-table label's were not aligning properly when data spanned large ranges. See merging PR for details.
- Fixed a bug in `find_best_parametric_model` where the wrong BIC value was being computed.
- Fixed regression bug when using an array as a penalizer in Cox models.

Bug fixes
- Fix integer-valued categorical variables in regression model predictions.
- numpy > 1.20 is allowed.
- Bug fix in the elastic-net penalty for Cox models that wasn't weighting the terms correctly.