FIX Preserve lifecycle on result persistence failures

[Roman Lutz (romanlutz)]

Description

Attack result persistence previously ran in a best-effort event handler. If the database write failed, the exception was ignored and the caller could receive a completed result even though no durable record existed for reporting or resume.

This change keeps generic lifecycle observers best-effort, enriches results during ON_POST_EXECUTE, and persists completed results only after teardown succeeds. Persistence failures now propagate without creating a misleading second ERROR result, while genuine execution or teardown failures continue to persist one accurate ERROR result and preserve the original exception.

N/A

Tests and Documentation

• Added regression coverage for post-teardown persistence failures, partial commits, duplicate prevention, teardown failures, and observer failures.
• Ran 61 focused strategy tests and 2,243 executor/scenario tests.
• Ran the changed-file pre-commit hooks, including Ruff and scoped type checking.
• Documentation changes were not applicable.

JupyText was not applicable because no documentation notebooks changed.

[hannahwestra25]

cp pointed out that this works if there's no error but if the attack & db write fail, the db error doesn't get saved.

so we'd need something like this:

try:
    result = await super().execute_with_context_async(context=context)
except Exception as attack_error:
    # build_error_result would use _on_error_async()
    error_result = self._default_event_handler._build_error_result(
        context=context,
        error=attack_error.__cause__ or attack_error,
    )
    try:
        self._default_event_handler._persist_result(result=error_result)
    except Exception as database_error:
        raise ExceptionGroup(
            "Attack and result persistence failed",
            [attack_error, database_error],
        )
    raise

self._default_event_handler._persist_result(result=result)
return result