Fix: Environment variables, rate limiting, and endoftext token issues

- Fix API key environment variables not being picked up from CLI
- Add retry logic with exponential backoff for rate limiting (2s, 4s, 8s delays)
- Add text sanitization to remove problematic tokens like <|endoftext|>
- Add configurable chunk delay (--chunk-delay) to avoid rate limits
- Increase default chunk delay from 2s to 5s
- Fix guidelines file path resolution for Cline rules
- Improve error handling and user feedback

Author: rainmana

lib/index.js

@@ -6,5 +6,6 @@ export interface LLMGeneratorOptions {
     maxTokens?: number;
     temperature?: number;
     chunkSize?: number;
+    chunkDelay?: number;
 }
 export declare function generateWithLLM(repoContent: string, guidelines: string, outputDir?: string, description?: string, ruleType?: string, outputFormat?: string, options?: LLMGeneratorOptions): Promise<string>;