Summary

Addresses the first suggestion in #902: pipeline.run required an output_name, even when the pipeline has a unique leaf node, while pipeline(...) (i.e., __call__) did not.

• Pipeline.run(output_name=None) (now the default) resolves to the unique leaf node's output name, or to all leaf nodes when there are multiple, returning a tuple of their outputs (as requested in Improvements / suggestions by #902: "just run the full pipeline, even without unique leaf nodes").
• Pipeline.__call__ now delegates the None case to run, so pipeline(a=1, b=2) also works for pipelines with multiple leaf nodes instead of raising.

Example

@pipefunc(output_name="c")
def f(a, b):
    return a + b

@pipefunc(output_name="d")
def g(c):
    return 2 * c

pipeline = Pipeline([f, g])
pipeline.run(kwargs={"a": 1, "b": 2})  # 6, no output_name needed

Verification

• New tests: unique leaf default, multiple-leaves default (tuple result, both via run and __call__), and a leaf with a multi-output output_name tuple.
• Full suite: 1396 passed, 11 skipped, 2 xfailed.
• pre-commit (ruff, mypy, etc.) passes.

Closes nothing on its own; part of #902.