Major projects submitted from different places share images

If two people with different databases submit major projects to the same s3 bucket, they get combined. Additionally, the way it works right now is it scans the s3 folder for images associated with the project, which is slow. Both of these problems could be solved by tracking images in the database so that displaying a project just displays the images that are actually associated with the major project.

Start by tracing the major-project submission and display flows, then inspect the existing S3 image scan and database models. Define how image associations are recorded for each project and verify that projects from different databases or submissions no longer mix, while display no longer scans the S3 folder.

由索引模型根据 Issue 内容生成。