Fix RANSAC for invalid model estimation and confidence corner case

Previously, estimators did not return whether the model estimation
was successful. RANSAC now tests whether the estimation was
successful and skips invalid models.

When the confidence/stop_probability of RANSAC was set to 1,
the iteration was falsely terminated early instead of running for
the maximum number of iterations.

skimage/transform/_geometric.py

@@ -263,6 +263,11 @@ class ProjectiveTransform(GeometricTransform):
         dst : (N, 2) array
             Destination coordinates.
 
+        Returns
+        -------
+        success : bool
+            True, if model estimation succeeds.
+
         """
 
         try:
@@ -270,7 +275,7 @@ class ProjectiveTransform(GeometricTransform):
             dst_matrix, dst = _center_and_normalize_points(dst)
         except ZeroDivisionError:
             self.params = np.nan * np.empty((3, 3))
-            return
+            return False
 
         xs = src[:, 0]
         ys = src[:, 1]
@@ -309,6 +314,8 @@ class ProjectiveTransform(GeometricTransform):
 
         self.params = H
 
+        return True
+
     def __add__(self, other):
         """Combine this transformation with another.
 
@@ -459,6 +466,11 @@ class PiecewiseAffineTransform(GeometricTransform):
         dst : (N, 2) array
             Destination coordinates.
 
+        Returns
+        -------
+        success : bool
+            True, if model estimation succeeds.
+
         """
 
         # forward piecewise affine
@@ -481,6 +493,8 @@ class PiecewiseAffineTransform(GeometricTransform):
             affine.estimate(dst[tri, :], src[tri, :])
             self.inverse_affines.append(affine)
 
+        return True
+
     def __call__(self, coords):
         """Apply forward transformation.
 
@@ -658,6 +672,11 @@ class SimilarityTransform(ProjectiveTransform):
         dst : (N, 2) array
             Destination coordinates.
 
+        Returns
+        -------
+        success : bool
+            True, if model estimation succeeds.
+
         """
 
         try:
@@ -665,7 +684,7 @@ class SimilarityTransform(ProjectiveTransform):
             dst_matrix, dst = _center_and_normalize_points(dst)
         except ZeroDivisionError:
             self.params = np.nan * np.empty((3, 3))
-            return
+            return False
 
         xs = src[:, 0]
         ys = src[:, 1]
@@ -699,6 +718,7 @@ class SimilarityTransform(ProjectiveTransform):
 
         self.params = S
 
+        return True
 
     @property
     def scale(self):
@@ -798,6 +818,11 @@ class PolynomialTransform(GeometricTransform):
         order : int, optional
             Polynomial order (number of coefficients is order + 1).
 
+        Returns
+        -------
+        success : bool
+            True, if model estimation succeeds.
+
         """
         xs = src[:, 0]
         ys = src[:, 1]
@@ -828,6 +853,8 @@ class PolynomialTransform(GeometricTransform):
 
         self.params = params.reshape((2, u // 2))
 
+        return True
+
     def __call__(self, coords):
         """Apply forward transformation.