fix: reject string-like shell commands

[Aphroq]

Summary

This pull request fixes shell call command normalization when the model returns a string-like action.commands payload.

• reject top-level str, bytes, and bytearray values in coerce_shell_call() instead of treating them as generic sequences
• raise a clear ModelBehaviorError before malformed payloads can be split into character-by-character commands
• add regression coverage for string-like commands payloads while keeping valid list-based shell calls unchanged

Test plan

• uv run pytest tests/test_shell_call_serialization.py
• bash .agents/skills/code-change-verification/scripts/run.sh

Issue number

Closes #3091

Checks

• I've added new tests (if relevant)
• I've added/updated the relevant documentation
• I've run make lint and make format
• I've made sure tests pass

[seratch]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Chef's kiss.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".