be explicit about task feature in TL (#2918)

Summary:
Pull Request resolved: #2918

X-link: facebook/Ax#4021

Previously, the TL adapter added a task feature to `SearchSpaceDigest.task_features`, but not to `SearchSpaceDigest.bounds` or `SearchSpaceDigest.feature_names`. This avoided needing to pass the task feature as a fixed_feature when making predictions or generating.

However, there was a bug when using `Normalize` that dropped one non-task feature here: https://github.com/facebook/Ax/blob/main/ax/generators/torch/botorch_modular/surrogate.py#L247-L250

This diff adds the task feature to via a new Transform. This significantly simplifies task feature handling across Adapter methods.

Reviewed By: saitcakmak

Differential Revision: D77905821

Author: sdaulton

botorch/utils/datasets.py

@@ -76,8 +76,8 @@ def __init__(
         self._X = X
         self._Y = Y
         self._Yvar = Yvar
-        self.feature_names = feature_names
-        self.outcome_names = outcome_names
+        self.feature_names = feature_names.copy()
+        self.outcome_names = outcome_names.copy()
         self.group_indices = group_indices
         self.validate_init = validate_init
         if validate_init:
@@ -351,7 +351,7 @@ def __init__(
         self.target_outcome_name = target_outcome_name
         self.task_feature_index = task_feature_index
         self._validate_datasets(datasets=datasets)
-        self.feature_names = self.datasets[target_outcome_name].feature_names
+        self.feature_names = self.datasets[target_outcome_name].feature_names.copy()
         self.outcome_names = [target_outcome_name]
 
         # Check if the datasets have identical feature sets.

test/utils/test_datasets.py

@@ -235,8 +235,8 @@ def test_clone(self, supervised: bool = True) -> None:
                     self.assertIs(dataset._X, dataset2._X)
                     self.assertIs(dataset._Y, dataset2._Y)
                     self.assertIs(dataset._Yvar, dataset2._Yvar)
-                    self.assertIs(dataset.feature_names, dataset2.feature_names)
-                    self.assertIs(dataset.outcome_names, dataset2.outcome_names)
+                    self.assertEqual(dataset.feature_names, dataset2.feature_names)
+                    self.assertEqual(dataset.outcome_names, dataset2.outcome_names)
                 # test with mask
                 mask = torch.tensor([0, 1, 1], dtype=torch.bool)
                 if supervised: