Pytorch-calibrated

Update enums to be strings instead of ints for easier use (e.g. use string instead of enum)

Fix: llvmlite version issue -- downgrade to 0.39.1 for installs to work
Fix: downgrade numpy version to 1.23.5 so the package works with shap (needs np.long)
Fix: remove python 3.11 support since it downgraded version of llvmlite requires <3.11

Initial PyTorch Calibrated Pre-Release Alpha

layers:
- NumericalCalibrator for calibrating numerical features. Supports bound and monotonicity constraints.
- CategoricalCalibrator for calibrating categorical features. Supports bound and pairwise monotonicity constraints.
- Linear for linearly combining inputs. Supports monotonicity and weighted average constraints.

configs:
- NumericalFeatureConfig for configuring numerical features for easy modeling using model constructors.
- CategoricalFeatureConfig for configuring categorical features for easy modeling using model constructors.

data:
- CSVData class for loading, preparing, and batching data for calibrated modeling.

enums:
- Monotonicity for specifying monotonicity constraints.
- NumericalCalibratorInit for specifying the initialization scheme to use for a NumericalCalibrator output keypoints.
- CategoricalCalibratorInit for specifying the initialization scheme to use for a CategoricalCalibrator output keypoints.
- InputKeypointsInit for specifying the initialization scheme to use for NumericalCalibrator input keypoints.
- FeatureType for determining the type of feature from its config.

models:
- CalibratedLinear model for easy construction of a calibrated linear model using feature configs.