feat: upgrade @huggingface/transformers to 3.7.2

- Memory optimizations: LRU cache for BPE tokenizer
- Performance improvements: optimized tensor.slice() method
- Enhanced quantization support for fp16/q8/q4
- Bug fixes: error handling, WebWorker detection, tokenizer padding
- No breaking changes, full backward compatibility maintained

package-lock.json

@@ -10,7 +10,7 @@
       "license": "MIT",
       "dependencies": {
         "@aws-sdk/client-s3": "^3.540.0",
-        "@huggingface/transformers": "^3.1.0",
+        "@huggingface/transformers": "^3.7.2",
         "boxen": "^8.0.1",
         "chalk": "^5.3.0",
         "cli-table3": "^0.6.5",