feat: add dynamic x-api-provider and reasoning_effort support

- Add per-model provider configuration in config.json
- Implement getModelProvider() to fetch provider from model config
- Update all header generators to accept dynamic provider parameter
- Add reasoning_effort field handling for common endpoint type
- Support auto/low/medium/high/off reasoning levels for common models

This enables flexible multi-provider support and reasoning control
across different endpoint types (anthropic, openai, common).

config.js

@@ -64,6 +64,11 @@ export function getModelReasoning(modelId) {
   return null;
 }
 
+export function getModelProvider(modelId) {
+  const model = getModelById(modelId);
+  return model?.provider || null;
+}
+
 export function getUserAgent() {
   return getCurrentUserAgent();
 }