Problem

When collecting tests that have parametrized values containing escaped unicode (like "\\u2603"), rtest generates different test IDs than pytest, causing pytest to fail with "not found" errors when using --runner pytest.

Reproduction

Given this test:

@pytest.mark.parametrize(
    "test_value,expected", [(True, '"\\u2603"'), (False, '"\u2603"')]
)
def test_json_as_unicode(test_value, expected, app, app_ctx):
    app.json.ensure_ascii = test_value
    rv = app.json.dumps("\N{SNOWMAN}")
    assert rv == expected

pytest generates test IDs:

• test_json_as_unicode[True-"\\u2603"]
• test_json_as_unicode[False-"\u2603"]

But rtest generates different IDs (likely without the proper escaping), causing:

ERROR: not found: tests/test_json.py::test_json_as_unicode
(no match in any of [<Module test_json.py>])

This happens on real-world projects like Flask's tests/test_json.py.

Context

The issue seems to be in how we serialize string literals containing backslash escapes when generating test IDs. The source string '"\\u2603"' contains a literal backslash followed by u2603, which pytest preserves verbatim in the test ID, but rtest may be handling the escaping differently.