Support partial/resumable downloads

[tecosaur]

When downloading something large, like an 80GB file, it's a massive pain when 70% of the way through the download the network stutters and you loose ... everything.

libcurl supports partial/resumable downloads (see https://curl.se/libcurl/c/CURLOPT_RESUME_FROM_LARGE.html), and so it would be nice if Downloads could support writing the partial download to a file/buffer and then continuing either from a manually specified point or just the size of the file/buffer so far.

Ideally, Downloads would also support automatically re-trying up to N times.

[StefanKarpinski]

I think the retry would be the way. Otherwise you need to introduce a whole API for providing a potentially partially downloaded file to continue. If the retrying is internal then that can be managed transparently.

[tecosaur]

Retrying is certainly the most "simple" way in many respects, however, I'm beginning to use Julia to manage large data sets (as I type this, I'm downloading a 750GB file, but via curl so I can resume it in case something goes wrong) and I'd imagine that just "retry" would fall short in cases like this.

Would the following sound viable?

• Add an option to dump partial output to a file
• Add an option to resume the download from a certain point

I'd think that would be sufficient to enable partially downloaded files.