Fix PaperBench code-only fallback for non-code rubrics

[sylvesterkaczmarek]

Summary

• make the code_only=True fallback produce a valid single Code Development task when a rubric contains no Code Development leaves
• avoid constructing invalid intermediate TaskNode states
• preserve the existing pruning path when Code Development criteria are present

run_judge() currently falls back with:

task_tree.set_task_category("Code Development").set_sub_tasks([])

when task_tree.code_only() returns None. TaskNode validation forbids a task category on an internal node, so the first call raises before set_sub_tasks([]) can run. Reversing the calls has the analogous problem because a leaf with no task category is also invalid.

The fix performs the fallback atomically with dataclasses.replace, creating the intended leaf without passing through an invalid intermediate tree.

Regression coverage verifies both the no-code fallback and the existing path where a Code Development subtree is present.