fix: block disabled function tools before execution

[Aphroq]

Summary

This change prevents agent-configured FunctionTools from executing when their dynamic is_enabled callback becomes false after the tool was exposed to the model but before execution starts.

The function tool executor now re-checks the currently enabled agent function tools before creating tool tasks. If the model called an agent-configured tool that is currently disabled, the run fails fast with ModelBehaviorError instead of invoking the tool. The check uses object identity for agent-configured tools so synthetic or adapter-provided FunctionTools that are not present in agent.tools can still execute.

This keeps is_enabled aligned with its documented runtime gating semantics for permissions, feature flags, quotas, tenant policy, and other side-effect-sensitive tool access.

Test plan

• uv run pytest tests/test_run_step_execution.py -k "disabled_before_execution or non_agent_function_tool" -q
• uv run pytest tests/test_run_step_execution.py -q
• bash .agents/skills/code-change-verification/scripts/run.sh

Issue number

Closes #3115

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".