Improve file update performance

[gqgs]

S3 doesn't allow partial updates to objects.
That means if we want to update only a small section of the object the only choice currently available is to download the full object, update the desired section, and write the full object again, overwriting the previous version.

This is clearly wasteful and inefficient.
An option to improve this, for some use cases (e.g. music), would be to split the metadata and actual data into different file objects.
This would allow a fast and efficient update of the file metadata.
Small downsides:

• Increase of complexity of managing files
• Additional get needed to get metadata (metadata + data gets can likely be done concurrently)