Use fuse/HTTP to parallelise snapshot download and import.

[elric1]

Rather than downloading and then importing a standard snapshot, we should be able to use one of the FUSE filesystems that present an HTTP file in the local filesystem. This way we wouldn't have to teach tezos-node about HTTP. I looked at one of them, but it was documented to have poor performance. It's worth a little research to see if one could find a better one and make it work.

[nicolasochem]

It's good, but having snapshot import from stdin would be better. (or snapshot import from url, so we don't even need curl)