Remove large cache files and update .gitignore to prevent future large file commits

Update model to cl-nagoya-sup-simcse-ja-nss-v1_0_8_4, add new standard name mapping functionality in prediction process, and enhance data loading in sentence transformer service.

Remove outdated anchor name sentence embeddings file, update cached embeddings with new sizes, and refactor sub subject and name mapping in the prediction process.

Update model to cl-nagoya-sup-simcse-ja-nss-v1_0_8_3, add new data files, and enhance startup process with directory creation and model loading.