feat: add approval override argument parity for RunState resumes

[seratch]

This pull request adds TypeScript-parity approval argument overrides to the Python SDK's HITL resume flow (see openai/openai-agents-js#1098 for TS implementation). It extends RunState.approve() with durable and execution-only override modes, persists the new override bookkeeping in schema 1.7, and updates replay surfaces so resumed tool execution, serialized state, and local canonical history stay aligned.

It also introduces Python equivalents of rewrite-aware and server-managed session capabilities. SQLiteSession and OpenAIResponsesCompactionSession can now rewrite persisted function-call history, OpenAIConversationsSession advertises server-managed history semantics for validation, and the runner now fails fast when a requested override mode is incompatible with the active session or conversation setup.

Finally, this pull request keeps compaction and tracing behavior consistent with the corrected transcript. Session persistence applies pending history mutations even on approval-only save cycles, compaction stays in input mode after local rewrites until a newer response id is observed, and approval overrides emit an audit custom span only when sensitive tracing is enabled. Regression coverage was added across sync resumes, streamed resumes, tracing, and compaction.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 1e8b70a05e

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".