[BUG] fix missing np.abs in ResidualDouble (#47)

This PR fixes an unreported bug, a missing `np.abs` in `ResidualDouble`.

Acknowledging that the estimator does not have all its features migrated
yet, but the abs operator was missing.

Tests did not catch this as this would just result in bad predictions.

skpro/regression/residual.py

@@ -4,6 +4,7 @@
 
 __author__ = ["fkiraly"]
 
+import numpy as np
 import pandas as pd
 from sklearn import clone
 
@@ -104,7 +105,7 @@ def _fit(self, X, y):
         y = y.values.flatten()
 
         est.fit(X, y)
-        resids = y - est.predict(X)
+        resids = np.abs(y - est.predict(X))
 
         resids = resids.flatten()