0.131.57

Fixed

• memo embeds and searches with the model chosen by memo models set. A downloaded EmbeddingGemma or
  e5-large was never used: memo embedded and searched with e5-small whatever was chosen, and embedding_model was
  not a memo setting. The next memo sync runs give your notes and mail vectors from the new model a part at a time;
  the old model's vectors are kept, so going back is instant.
• The missing-model message names the model that is missing and the exact download, with --accept-terms for a
  model that needs it; it always named e5-small.
• A sync with a slower model embeds fewer chunks a run, so a timer run still ends within its five minutes.
• memo doctor reports the search model in use, not always e5-small.