Do not store exported content hashes as a file

When I implemented the content hashing feature, I thought that committing hashes to Git would be a big performance improvement, but never ran performance testing. After just testing with 100, 1000, and 10000 entities, I think the performance difference is negligible and want to change this before releasing it in the wild. We should just use sha1_file on every export instead, and should probably change the \Drupal\tome_sync\ContentHasherInterface API so that getChangedContent/getDeletedContent/getNewContent all come from one method, to avoid doing this multiple times.