[rollout_trace] Add rollout trace crate

[cassirer-openai]

Summary

Adds the standalone codex-rollout-trace crate, which defines the raw trace event format, replay/reduction model, writer, and reducer logic for reconstructing model-visible conversation/runtime state from recorded rollout data.

The crate-level design is documented in codex-rs/rollout-trace/README.md.

Stack

This is PR 1/5 in the rollout trace stack.

• #18876: Add rollout trace crate
• #18877: Record core session rollout traces
• #18878: Trace tool and code-mode boundaries
• #18879: Trace sessions and multi-agent edges
• #18880: Add debug trace reduction command

Review Notes

This PR intentionally does not wire tracing into live Codex execution. It establishes the data model and reducer contract first, with crate-local tests covering conversation reconstruction, compaction boundaries, tool/session edges, and code-cell lifecycle reduction. Later PRs emit into this model.

The README is the best entry point for reviewing the intended trace format and reduction semantics before diving into the reducer modules.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f1f33a055d

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".

[jif-oai]

The shape looks good but this PR implies that the producer needs to build the objects (e.g. rawTraceEventPayload::ToolCallStarted from the tool dispatch). I think it would be better to provide helpers to the producer such that they don't get too polluted and they don't have to know about the details of the struct

We could imagine something like this:

  let tool_trace = ToolTraceContext::enabled(
      trace_writer.clone(),
      thread_id,
      codex_turn_id,
  );

  let attempt = tool_trace.start(ToolTraceStart {
      tool_call_id,
      caller: ToolCaller::CodeCell { runtime_cell_id },
      model_call_id,
      dispatch: &dispatch_request,
  });

  attempt.record_runtime_started(&protocol_begin_event);
  attempt.record_runtime_ended(status, &protocol_end_event);
  attempt.record_completed(status, &tool_result);

[jif-oai]

Let's consider it as a basis and move on. Can you just add a README explaining we won't collect users data? We are OSS and we will get some pings otherwise

[cassirer-openai]

The README already says it is opt in but I made it even more clearer that nothing leaves the machine and that has to be explicitly enabled to do anything.