config: Use append mode for LanceDB storage

Langchain, in their infinite wisdom, default to the `overwrite` mode.
This means that every time you store _any_ document or text in LanceDB,
the _entire table_ is erased.  This does not make sense for our use case.

Closes: AP-503

Author: awilfox

willa/config/__init__.py

@@ -126,7 +126,8 @@ def get_lance() -> LanceDB:
                                                   base_url=CONFIG['OLLAMA_URL'])
     else:  # If we add another backend, elif CONFIG['EMBED_BACKEND'] == 'bedrock':
         embeddings = BedrockEmbeddings(model_id=CONFIG['EMBED_MODEL'], provider='cohere')
-    return LanceDB(embedding=embeddings, uri=CONFIG['LANCEDB_URI'], table_name='willa')
+    return LanceDB(embedding=embeddings, uri=CONFIG['LANCEDB_URI'], table_name='willa',
+                   mode='append')
 
 
 def get_model() -> BaseChatModel: