[ML] Accept a close match for results of DFA evaluation tests

The results of these tests can vary in the last significant digit
depending on hardware and maths library optimisations.

Therefore we need to tolerate small failures in results to avoid
spurious failures.

x-pack/plugin/src/yamlRestTest/resources/rest-api-spec/test/ml/evaluate_data_frame.yml

@@ -705,6 +705,8 @@ setup:
           }
 ---
 "Test classification auc_roc":
+  - skip:
+      features: close_to
   - do:
       ml.evaluate_data_frame:
         body:  >
@@ -722,10 +724,12 @@ setup:
               }
             }
           }
-  - match: { classification.auc_roc.value: 0.8612749970197677 }
+  - close_to: { classification.auc_roc.value: { value: 0.8612749970197677, error: 0.00000000001 }}
   - is_false: classification.auc_roc.curve
 ---
 "Test classification auc_roc with default top_classes_field":
+  - skip:
+      features: close_to
   - do:
       ml.evaluate_data_frame:
         body:  >
@@ -742,7 +746,7 @@ setup:
               }
             }
           }
-  - match: { classification.auc_roc.value: 0.8612749970197677 }
+  - close_to: { classification.auc_roc.value: { value: 0.8612749970197677, error: 0.00000000001 }}
   - is_false: classification.auc_roc.curve
 ---
 "Test classification accuracy with missing predicted_field":
@@ -1014,6 +1018,8 @@ setup:
           }
 ---
 "Test regression mean_squared_error":
+  - skip:
+      features: close_to
   - do:
       ml.evaluate_data_frame:
         body: >
@@ -1028,12 +1034,14 @@ setup:
             }
           }
 
-  - match: { regression.mse.value: 28.67749840974834 }
+  - close_to: { regression.mse.value: { value: 28.67749840974834, error: 0.00000000001 }}
   - is_false: regression.msle.value
   - is_false: regression.r_squared.value
   - is_false: regression.huber.value
 ---
 "Test regression mean_squared_logarithmic_error":
+  - skip:
+      features: close_to
   - do:
       ml.evaluate_data_frame:
         body: >
@@ -1048,12 +1056,14 @@ setup:
             }
           }
 
-  - match: { regression.msle.value: 0.08680568028334916 }
+  - close_to: { regression.msle.value: { value: 0.08680568028334916, error: 0.00000000001 }}
   - is_false: regression.mse.value
   - is_false: regression.r_squared.value
   - is_false: regression.huber.value
 ---
 "Test regression huber":
+  - skip:
+      features: close_to
   - do:
       ml.evaluate_data_frame:
         body: >
@@ -1068,12 +1078,14 @@ setup:
             }
           }
 
-  - match: { regression.huber.value: 3.5088110471730145 }
+  - close_to: { regression.huber.value: { value: 3.5088110471730145, error: 0.00000000001 }}
   - is_false: regression.msle.value
   - is_false: regression.mse.value
   - is_false: regression.r_squared.value
 ---
 "Test regression r_squared":
+  - skip:
+      features: close_to
   - do:
       ml.evaluate_data_frame:
         body: >
@@ -1087,13 +1099,15 @@ setup:
               }
             }
           }
-  - match: { regression.r_squared.value: 0.8551031778603486 }
+  - close_to: { regression.r_squared.value: { value: 0.8551031778603486, error: 0.00000000001 }}
   - is_false: regression.mse
   - is_false: regression.msle.value
   - is_false: regression.huber.value
 
 ---
 "Test regression with null metrics":
+  - skip:
+      features: close_to
   - do:
       ml.evaluate_data_frame:
         body: >
@@ -1107,9 +1121,9 @@ setup:
             }
           }
 
-  - match: { regression.mse.value: 28.67749840974834 }
-  - match: { regression.r_squared.value: 0.8551031778603486 }
-  - match: { regression.huber.value: 1.9205280586939963 }
+  - close_to: { regression.mse.value: { value: 28.67749840974834, error: 0.00000000001 }}
+  - close_to: { regression.r_squared.value: { value: 0.8551031778603486, error: 0.00000000001 }}
+  - close_to: { regression.huber.value: { value: 1.9205280586939963, error: 0.00000000001 }}
   - is_false: regression.msle.value
 ---
 "Test regression given missing actual_field":