[Bug] Status bar shows base thinking effort instead of forced effort from KIMI_MODEL_THINKING_EFFORT

Summary

When KIMI_MODEL_THINKING_EFFORT=max is set, the actual LLM requests correctly run at max effort, but the TUI status bar (footer) still displays the base effort from config.toml (thinking: high). Display and effective behavior disagree.

Environment

• Kimi Code version: 0.43.1 (macOS arm64)
• Provider: managed:kimi-code (type kimi)
• Model: kimi-code/k3-256k (support_efforts = ["low", "high", "max"])
• config.toml: [thinking] enabled = true, effort = "high"

Steps to reproduce

KIMI_MODEL_THINKING_EFFORT=max kimi --yolo

Expected behavior

The footer status bar shows thinking: max, matching the effort actually used for requests.

Actual behavior

Footer shows thinking: high (the [thinking].effort value from config.toml).

Evidence

The env var does take effect for real requests — a session run with it set logs:

thinkingEffort=max systemPromptChars=39066 toolCount=27

in the session log, while the same session's profile.bind wire event records "thinkingEffort":"high", which is what the status bar renders.

From the shipped binary, the footer appears to render the profile's base thinkingLevel, while the env var is bound as [thinking].forcedEffort and only layered on at request time via resolveForcedThinkingEffort(...) (effective: forced ?? base). So the env override never reaches the status-bar rendering path.

Suggested fix

Have the status bar display the effective thinking effort (after applying forcedEffort), not just the base profile level — or otherwise surface that a forced effort is active.

Locate the status-bar rendering path that uses the profile's base thinkingLevel, then compare it with resolveForcedThinkingEffort(...) and the request-time effective value. Reproduce with KIMI_MODEL_THINKING_EFFORT=max kimi --yolo and verify that the footer shows thinking: max while requests still use max effort.