add support `xgboost` models

[isabelizimm]

Describe the solution you'd like
Request for new handler for xgboost package

Additional context
XGBoost is currently supported by vetiver-r, and is a multilingual package. This would be a great step for parity between R and Python vetiver.

[gsingh91]

Another model type we can/should support is tensorflow. Native Tensorflow support on Connect is planned to be phased out, so a way to host it through vetiver would be a good alternative.

[machow]

Looking at XGBoost models, I'm noticing that some inherit from sklearn (e.g. xgboost.XGBClassifier, XGBModel). If we need to support these models, it seems like there are two pieces to figure out:

1. can the models that inherit from sklearn classes just use the sklearn handler? It seems like the big piece here is whether these classes also require the input be wrapped in xgboost.DMatrix. In that case we'll probably need to subclass the sklearn handler!
2. what other handlers need to be registered? (maybe one for xgboost.core.Booster?)

XGBClassifier:

In [8]: xgboost.XGBClassifier.mro()
Out[8]:
[xgboost.sklearn.XGBClassifier,
 xgboost.sklearn.XGBModel,
 sklearn.base.BaseEstimator,
 sklearn.base.ClassifierMixin,
 object]

Booster, which is afaik returned by xgboost.predict:

In [11]: xgboost.core.Booster.mro()
Out[11]: [xgboost.core.Booster, object]