Summary
Create a repository of corrupted databases, ideally of different kinds (the ones created from different kinds of issues seen in the field.) This should help us in the future to train the team to debug, also give us an opportunity to create tools to discover and debug these corruptions.

This is motivated by PM-2569 (see the document here).

Acceptance Criteria (Definition of Done)
I think we should find one or more recent crashes, that show some kind of corruption in a WT file and preserve the errant file or files probably in some (new?) designated S3 repository. The files for a single fail should be stored in a separate directory designating the failure (perhaps with a README describing what is wrong, the date, ticket number, etc.). Then, write a simple evergreen test, perhaps to run salvage on the file and confirm that the repository can be accessed and to show that expected contents have been recovered.

Once this is in place, we can close this ticket knowing that a basic repository is set up. From there, we'll need to use our imagination to see how the files could be used. These could be educational (learn how to debug/diagnose/fix corruption failures), or test new tools that run on files, or existing tools, like verify/salvage.