[ENH] multiple quantile regression (#108)

Fixes #107

For quantile regression, often more than one quantile probability is of
interest. However, existing Sklearn compatible quantile regressors
always fit and predict a single quantile probability. To the best of my
knowdlegde, there is no standardized way to integrate multiple quantile
regression with Sktime/Skpro probabilistic prediction methods such as
predict_quantiles/predict_intervals. This PR adds new Skpro regressor
that wraps multiple quantile regressors and supports probabilistic
predictions from wrapped regressors.

.all-contributorsrc

@@ -88,6 +88,17 @@
         "ideas",
         "review"
       ]
+    },
+    {
+      "login": "Ram0nB",
+      "name": "Ramon Bussing",
+      "avatar_url": "https://avatars.githubusercontent.com/u/45173421",
+      "profile": "https://github.com/Ram0nB",
+      "contributions": [
+        "code",
+        "doc",
+        "ideas"
+      ]
     }
   ]
 }

docs/source/api_reference/regression.rst

@@ -63,6 +63,14 @@ take one or multiple ``sklearn`` estimators and adda probabilistic prediction mo
 
     ResidualDouble
 
+.. currentmodule:: skpro.regression.multiquantile
+
+.. autosummary::
+    :toctree: auto_generated/
+    :template: class.rst
+
+    MultipleQuantileRegressor
+
 Base
 ----