feat: add support for Claude LLM integration across multiple modules

- Introduced `useClaude` option in `AnalysisPipelineOptions` to toggle Claude LLM usage.
- Updated `processProductChunk` and `analyzeProducts` functions to accept and handle `useClaude` parameter.
- Modified argument parsing in various scripts (`bestsellers-by-category`, `mid-range-sellers-by-category`, `top-monthly-sold-by-category`, etc.) to include `--claude` flag.
- Enhanced `analyzeProductsInternal` to differentiate between LLM providers and handle requests to Claude API.
- Added error handling for Claude API responses and ensured proper configuration for using Claude.
- Updated documentation and usage messages to reflect the new `--claude` flag.

src/analysis-pipeline.ts

@@ -22,6 +22,7 @@ export type AnalysisPipelineOptions = {
   llmBatchDelayMs?: number;
   llmRetryDelayMs?: number;
   sellability?: SellabilityFilter;
+  useClaude?: boolean;
 };
 
 export function chunkArray<T>(items: T[], chunkSize: number): T[][] {
@@ -60,6 +61,7 @@ export async function processProductChunk(
   const llmBatchDelayMs = Math.max(0, options.llmBatchDelayMs ?? 5_000);
   const llmRetryDelayMs = Math.max(0, options.llmRetryDelayMs ?? 10_000);
   const sellabilityFilter = options.sellability ?? "available";
+  const useClaude = options.useClaude === true;
 
   console.log(`\nChecking cache for ${products.length} products...`);
   const cached = new Map<string, EnrichedProduct>();
@@ -242,6 +244,7 @@ export async function processProductChunk(
     try {
       verdicts = await analyzeProducts(batch, {
         ignoreSellability: sellabilityFilter === "all",
+        useClaude,
       });
     } catch {
       if (llmRetryDelayMs > 0) {
@@ -250,6 +253,7 @@ export async function processProductChunk(
       try {
         verdicts = await analyzeProducts(batch, {
           ignoreSellability: sellabilityFilter === "all",
+          useClaude,
         });
       } catch {
         verdicts = null;