OpenPechaS3 class

[eroux]

It seems our current workflow won't scale very well because of Github limitations (see this example). An alternative is to store the opf files on an s3 bucket. The bucket name could be opf.openpecha.org and there could be an intermediate step of folders like what we have on archive.tbrc.org or ocr.bdrc.io, using the last two letters of the pecha id for instance. Once it's done we should copy the opfs currently on Github on s3.

[eroux]

Now, obviously, the issue is that these won't be git repositories anymore. Some ideas:

• perhaps we don't always need git repositories, but I think we should keep git repos
• we could store the git repos as bare git repositories (there's already a class for that but it could be improved), this will lower the storage cost but will be a little less convenient. Since we'll replace the files after each change, conflicts, etc. won't be handled but I think we can live with that