[BUG] `CompilePipeline` adds duplicate expand transform when called on a `QNode`

[andrijapau]

Expected behavior

The pipeline of the QNode should not duplicate any expand transforms.

Actual behavior

When the pipeline is created, we automatically add the appropriate expand transform due to metric_tensor. However, when the pipeline is being called on the QNode, and we are processing it transform by transform, we end up adding another expand transform when processing metric_tensor.

Additional information

No response

Source code

pipeline = qml.transforms.merge_rotations + qml.metric_tensor

@pipeline
@qml.qnode(qml.device("default.qubit", wires=2))
def circuit():
    return qml.expval(qml.Z(0))

print(circuit.compile_pipeline)

Tracebacks

CompilePipeline(
  [0] merge_rotations(),
  [1] _expand_metric_tensor(), # DUPLICATE
  [2] _expand_metric_tensor(device_wires=Wires([0, 1])),
  [3] metric_tensor(device_wires=Wires([0, 1]))
)

System information

PL v0.44

Existing GitHub issues

• I have searched existing GitHub issues to make sure the issue does not already exist.