Fall back to repr() when __pt_repr__ yields malformed fragments

[chiliec]

Fixes #610.

Problem

Inspecting a unittest.mock.MagicMock result crashes the REPL:

>>> from unittest.mock import MagicMock
>>> m = MagicMock()
>>> m.get()
>>> m.method_calls[0]
Traceback (most recent call last):
  ...
  File ".../prompt_toolkit/formatted_text/utils.py", line 88, in split_lines
    for style, string, *mouse_handler in fragments:
ValueError: not enough values to unpack (expected at least 2, got 0)

Root cause

In OutputPrinter._format_result_output, a result carrying __pt_repr__ is trusted to return prompt_toolkit formatted text. A MagicMock fabricates any attribute, so hasattr(result, "__pt_repr__") is True; calling it returns further mocks, and to_formatted_text(...) then produces a malformed fragment list such as:

FormattedText(["__pt_repr__().__pt_formatted_text__", (), {}])

The empty tuple () has no (style, text) to unpack, so the later split_lines pass raises ValueError.

Fix

After building the formatted text from __pt_repr__, verify that every fragment is a well-formed (style, text, …) tuple before yielding it. If any fragment is malformed, fall through to the normal repr() path (which lexes call.get() into valid fragments). This keeps __pt_repr__ support intact for well-behaved objects and only guards against fabricated / broken implementations.

Test

Adds tests/test_printer.py, which feeds MagicMock().method_calls[0] through _format_result_output and asserts the fragments are all valid tuples, split_lines does not raise, and the rendered text is the object’s repr.

Verification (local, Python 3.11)

• RED→GREEN: with the fix reverted, the new test fails (the fabricated fragment is not a tuple); with the fix it passes.
• Full test suite passes (pytest), and ruff check on the changed files is clean.