codex doctor does not report the effective model_reasoning_effort

What variant of Codex are you using?

CLI (codex-cli 0.153.4)

What feature would you like to see?

codex doctor --json reports model and model provider under config.load, but
not model_reasoning_effort. Please add it.

There's currently no CLI surface that exposes the effective effort: doctor doesn't
have it, and codex debug prompt-input renders byte-identical output for
-c model_reasoning_effort=low and =high. So tooling has to parse
~/.codex/config.toml and redo profile/env resolution itself.

Looks like a one-liner in config_check() (codex-rs/cli/src/doctor.rs), next to
the existing model push. No "effort" string anywhere in that file on 0.153.4,
rust-v0.154.0-alpha.11, or main, so I don't think it's already in flight.

Additional information

Why I want it: resuming a thread under a different effort throws away its prompt
cache, so a wrapper doing exec + exec resume needs to check before resuming.

Two identical threads on 0.153.4, one-word prompts, only turn 2's effort differs
(turn.completed.usage):

72.8% is exactly what a fresh run gets, i.e. the thread's own history stops being
cached. Row 3 is that same thread one turn later with effort unchanged, so it's the
change that costs, not high effort itself. It scales with thread size: on a
600k-token thread that's ~590k tokens re-billed as uncached.

Repro: codex exec --json -c model_reasoning_effort=low --sandbox read-only -, then
resume that thread id twice, once with =low and once with =high, and compare
usage.cached_input_tokens.

Open codex-rs/cli/src/doctor.rs and inspect config_check(), starting next to the existing model push under config.load. Expose the effective model_reasoning_effort there, then verify that codex doctor --json reports it after configuration and profile/env resolution.