feat(codex): declare gpt-5.6 context window for Claude Code takeover

- Update Codex OAuth presets to the gpt-5.6 family (haiku -> gpt-5.6-luna)
  and bump the custom Codex template default model
- Inject CLAUDE_CODE_MAX_CONTEXT_TOKENS / CLAUDE_CODE_AUTO_COMPACT_WINDOW
  (372000, the ChatGPT Codex catalog window with a ~353K effective budget,
  openai/codex#31860) into effective live settings so Claude Code stops
  assuming a 200K window and compacts before the upstream rejects the prompt
- Gate the injected defaults on every configured model being gpt-5.6*:
  gpt-5.5's upstream catalog oscillates between 272K and 372K and must not
  inherit them; explicit user values always win
- Strip the injected values on switch-away backfill (mirror-inverse of the
  injection conditions) so program defaults never harden into per-provider
  explicit values, and keep both keys out of the shared common-config snippet

src-tauri/src/services/provider/mod.rs

@@ -722,6 +722,8 @@ mod tests {
                 "ANTHROPIC_BASE_URL": "https://example.com",
                 "ANTHROPIC_MODEL": "claude-x",
                 "CLAUDE_CODE_SUBAGENT_MODEL": "gpt-5.4-mini",
+                "CLAUDE_CODE_MAX_CONTEXT_TOKENS": "400000",
+                "CLAUDE_CODE_AUTO_COMPACT_WINDOW": "400000",
                 // 可共享、非机密配置（复数 _TOKENS 不应被误剥）
                 "ENABLE_TOOL_SEARCH": "true",
                 "CLAUDE_CODE_MAX_OUTPUT_TOKENS": "8192"
@@ -767,6 +769,12 @@ mod tests {
         assert!(env
             .and_then(|e| e.get("CLAUDE_CODE_SUBAGENT_MODEL"))
             .is_none());
+        assert!(env
+            .and_then(|e| e.get("CLAUDE_CODE_MAX_CONTEXT_TOKENS"))
+            .is_none());
+        assert!(env
+            .and_then(|e| e.get("CLAUDE_CODE_AUTO_COMPACT_WINDOW"))
+            .is_none());
 
         // 可共享的非机密配置必须保留（含复数 _TOKENS 不被误剥）
         assert_eq!(
@@ -2935,6 +2943,11 @@ impl ProviderService {
             "ANTHROPIC_DEFAULT_FABLE_MODEL",
             "ANTHROPIC_DEFAULT_FABLE_MODEL_NAME",
             "CLAUDE_CODE_SUBAGENT_MODEL",
+            // Context limits follow the actual upstream model. Sharing these
+            // across providers can cap GPT/Kimi to the wrong window and make
+            // Claude Code compact too early or miss the upstream limit.
+            "CLAUDE_CODE_MAX_CONTEXT_TOKENS",
+            "CLAUDE_CODE_AUTO_COMPACT_WINDOW",
             "ANTHROPIC_BASE_URL",
         ];