/usage: Z.AI coding-plan (API key) shows no real usage limits/balance, only key-status + local spend

Description
With a Z.AI coding-plan subscription configured via API key (login provider zai, OpenAI-compatible profile ZAI_PROFILE, base https://api.z.ai/api/coding/paas/v4), /usage does not show real usage limits or remaining quota. It only shows a free GET /models key-status probe plus locally tracked spend. The OAuth surface (Claude/OpenAI) shows full usage windows, and other coding agents (e.g. the Z.AI coding plan in other tools) show the actual remaining quota for this plan type.

Repro

1. Configure Z.AI coding-plan via API key (ZHIPU_API_KEY in zai.env).
2. Use the provider so it has activity.
3. Run /usage.
4. Z.AI (API key) row shows only key status and local spend — no plan usage/limit/remaining quota.

Expected
The Z.AI row shows the coding-plan usage/remaining quota the same way OAuth surfaces (Claude/OpenAI) do, or as other coding agents report it.

Root cause (from source)
In crates/jcode-base/src/usage/api_keys.rs, enqueue_api_key_usage_tasks only gives real balance fetchers to deepseek and moonshotai:

let has_balance_api = matches!(profile.id, "deepseek" | "moonshotai");

Everything else (including zai) falls into the generic branch in fetch_compatible_profile_report, which only does a free GET {base}/models probe for key validity and appends locally tracked spend. There is no Z.AI coding-plan usage/quota fetcher, so the coding-plan subscription data that the OAuth-style surfaces show is never surfaced for the API-key route.

Suggested area
Add a Z.AI coding-plan usage fetcher (e.g. quota/remaining-tokens endpoint under https://api.z.ai/api/coding/paas/v4) and register zai in has_balance_api / fetch_compatible_profile_report, matching how DeepSeek/Moonshot balance is wired today.

Start in crates/jcode-base/src/usage/api_keys.rs, reading enqueue_api_key_usage_tasks and fetch_compatible_profile_report alongside the existing DeepSeek and Moonshot balance handling. Investigate the Z.AI coding-plan quota or remaining-tokens endpoint under the configured API base, then wire the fetcher for the zai profile. Done means /usage shows Z.AI coding-plan usage, limits, or remaining quota rather than only key status and local spend.