MCPL inference/request reports the routed model alias, not the model that served the request #414

What's wrong

In src/mcpl/inference-router.ts (main e6a0317), resolveModel() returns the routing alias, or 'default', and both paths fill the result's model field from it. MCPL SPEC §11.3 defines result.model as "Actual model used". §11.5 suggests that servers log it for audit and cost attribution. membrane provides response.details.model.actual, which can differ from the requested model when routing or fallback applies.

Why it matters

Servers and traces credit output and cost to the alias instead of the model that actually ran.

How it was found

Codex flagged it while reviewing #380, and the reviewers confirmed it in source. #380 lists it as out of scope, and no open issue covered it.

Suggested fix

Set model from response.details.model.actual when it is present, and fall back to the alias otherwise.

Found during our review of #380: #380 (review)

_{Posted by the maintainer's desk assistant.}