Use redirect instead of download to use s3

[matthiasL-scality]

This pull request updates the NGINX configuration to improve how file downloads from the /builds/ path are handled, providing better support for both CLI and browser clients. The main change is the introduction of logic to distinguish between CLI tools and browsers, optimizing download behavior and maintaining compatibility for directory listings and metadata requests.

Improvements to download handling:

• Added logic to rewrite requests from CLI clients (like curl, wget, aria2c) to /redirect/, so they receive a presigned S3 URL and download files directly, while browser clients continue to be proxied through NGINX via /download/.
• Ensured that directory listings (requests with a trailing slash) always use the /download/ path, preserving the original behavior for browsers.
• Updated handling for HEAD requests so that these metadata checks are always proxied directly to /download/, regardless of client type.

[bertranddemiddelaer]

This one sounds like a good idea, but it will not be fully transparent for some GET implementations which do not support location header by default. Curl will need a '-L' for example. This may require a wrapper or an alias to add a '-L' in a transparent way for our beloved CI customers. The other option is to let them fix all their workflows and calls inside their scripts.

[bertranddemiddelaer]

LGTM

Beware of those changes however, they will break any CI workflow with artifacts curl calls without '-L'