fix the word embedding cache being included in git and try switching to v3 which might fix jina being so slow

.gitignore

@@ -180,4 +180,5 @@ test*.py
 logs/
 logs
 uploads/
-data.db
+data.db
+compared_word_emb.json

jina.py

@@ -14,7 +14,7 @@ def get_embedding(input: list):
     headers["Authorization"] = f"Bearer {os.environ['JINA_TOKEN']}"
 
     data = {
-        "model": "jina-embeddings-v4",
+        "model": "jina-embeddings-v3",
         "task": "text-matching",
         "input": input
     }