ci: Meta Llama - reduce parallelism to avoid rate limits (#2839)

Author: anakin87

.github/workflows/meta_llama.yml

@@ -33,7 +33,7 @@ jobs:
       matrix:
         os: [ubuntu-latest, windows-latest, macos-latest]
         python-version: ["3.10", "3.13"]
-      max-parallel: 3  # to avoid rate limiting        
+      max-parallel: 2  # to avoid rate limiting        
 
     steps:
       - name: Support longpaths