cerbero: Switch from wget to curl on Fedora 40 and newer

Fedora 40 switched from wget to wget2, and doesn't backport critical bugfixes such as returning a non-zero status code on error, which can happen due to any number of reasons, such as certificate incompatibilities, network issues, domain not existing, server issues, etc.

In these cases we get a success code and we don't try the fallback URL and we fail in extract() with a confusing error message saying that the downloaded tarball was not found.

Switch to curl, which doesn't have such issues.

I encountered this issue while building sqlite3, and I thought it was because the URL was gone upstream so I bumped it. Then I found that wget2 completely failed to connect to sqlite.org, at least on my system.