fix(models): fix builtin provider detection and model matching (#120)

- Add glm-coding-plan to PROVIDER_ENV_MAP for proper env mapping
- Rename GLMCodingPlan value from 'glm' to 'glm-coding-plan' (kebab-case)
- Match custom providers against PROVIDER_PRESETS to reuse builtin models
- Fix provider key matching in create/update (use entry.name consistently)
- Clear stale base_url/api_key from config on provider create
- Clear model config when all providers are removed
- Add gateway restart on provider remove

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

packages/server/src/services/config-helpers.ts

@@ -9,6 +9,7 @@ import { logger } from './logger'
 // --- Provider env var mapping (from hermes providers.py HERMES_OVERLAYS + config.py) ---
 export const PROVIDER_ENV_MAP: Record<string, { api_key_env: string; base_url_env: string }> = {
   openrouter: { api_key_env: 'OPENROUTER_API_KEY', base_url_env: '' },
+  'glm-coding-plan': { api_key_env: '', base_url_env: '' },
   zai: { api_key_env: 'GLM_API_KEY', base_url_env: '' },
   'kimi-coding-cn': { api_key_env: 'KIMI_CN_API_KEY', base_url_env: '' },
   moonshot: { api_key_env: 'MOONSHOT_API_KEY', base_url_env: '' },