#2969 Expose optional pricing in model listings for client usage estimates

说明

## Use case

I run Hermes and OpenCode against Strata and want to compare their usage using the same simulated external API rates. This lets me compare session costs in familiar units, even though inference runs locally.

Hermes can read pricing metadata from the endpoint's model listing. Strata currently omits it, so I cannot supply those rates through model discovery.

## Proposed change

Allow an optional top-level `pricing` object in the run config and expose it in `/v1/models` and `/models`, including alias entries.

The rates would follow OpenRouter's pricing convention: USD per token for prompt, completion, and cache usage, plus an optional per-request fee. These would be operator-supplied accounting rates; clients would calculate the estimates.

Existing configs without pricing would keep their current behavior and require no edits.

This relates to the usage dashboard work in #2204, #2370, and #2400, but only concerns exposing rates to external clients. I have a small implementation ready with validation, documentation, and tests.