Avoid IndexError when the pytest traceback cut has a single entry

[rachfop]

SybilItem._traceback_filter re-roots the displayed traceback at the frame
below the example frame (tb[1]) to hide Sybil's own frames. When the example
frame is the deepest frame in a failure's traceback, traceback.cut(path=example_module_path)
collapses to a single entry, so tb[1] raises IndexError.

Because this runs inside pytest's failure-reporting path, that IndexError
surfaces as a pytest INTERNALERROR that aborts the whole session and hides the
example that actually failed, instead of reporting it.

I hit this with a library whose examples execute through a deferred realization
step: the failure's traceback had the Sybil example frame as its deepest frame,
collapsing the cut to one entry.

The fix guards the lookup — when the cut has more than one entry the re-rooting
is unchanged; when it has a single entry the cut traceback is kept as-is, so the
failure is reported normally.

Includes a regression test in tests/test_functional.py that reproduces the
single-entry cut and asserts _traceback_filter no longer raises. ./happy.sh
passes (tests at 100% coverage, ruff, mypy, docs build).

[cjw296]

I hit this with a library whose examples execute through a deferred realization
step: the failure's traceback had the Sybil example frame as its deepest frame,
collapsing the cut to one entry.

Why is this framework truncating the stack frames of the traceback? The test you've added does this is an artificial way, but I'd like to understand what's happening in the real code.

[cjw296]

@rachfop - ping? Will close as stale in a week or so...

[cjw296]

@adamtheturtle - not sure if you have context on this, but closing as stale for now.

[adamtheturtle]

No ideas from me.