Make NoDimensionsTask test utility consistent about storage classes.

The config option provided by this task to change its output storage
class actually only change the Python type returned, making that type
inconsistent with the storage class it declared in its connections.
This means we've long been testing essentially the wrong behavior, and
I bet it's part of why we've had to be so defensive about storage
class conversions in butler.

This also switches the default for outputSC from 'dict' to
'StructuredDataDict', since it's supposed to be a storage class name,
not a pytype.  This wasn't causing trouble, but it was very confusing.

python/lsst/pipe/base/tests/no_dimensions.py

@@ -27,6 +27,7 @@
     "NoDimensionsTestTask",
 )
 
+import dataclasses
 from typing import cast
 
 from lsst.pex.config import Field
@@ -51,13 +52,19 @@ class NoDimensionsTestConnections(PipelineTaskConnections, dimensions=set()):
         name="output", doc="some dict-y output data for testing", storageClass="StructuredDataDict"
     )
 
+    config: NoDimensionsTestConfig
+
+    def __init__(self, *, config: PipelineTaskConfig | None = None):
+        if self.config.outputSC != "StructuredDataDict":
+            self.output = dataclasses.replace(self.output, storageClass=self.config.outputSC)
+
 
 class NoDimensionsTestConfig(PipelineTaskConfig, pipelineConnections=NoDimensionsTestConnections):
     """Configuration for `NoDimensionTestTask`."""
 
     key = Field[str](doc="String key for the dict entry the task sets.", default="one")
     value = Field[int](doc="Integer value for the dict entry the task sets.", default=1)
-    outputSC = Field[str](doc="Output storage class requested", default="dict")
+    outputSC = Field[str](doc="Output Python type requested", default="StructuredDataDict")
 
 
 class NoDimensionsTestTask(PipelineTask):