Redundant HTTP requests for the .zsync file

[DeeDeeG]

Hi,

As mentioned in #33...

the first 1024 bits of the remote .zsyncfile are requested twice, and then the full .zsync file is potentially requested twice as well.

[DeeDeeG]

Here's a more detailed breakdown of this bug.

For background, I think the program flow is something like this:

• Grab first KB of .zsync file, which includes some useful metadata.
• Read the pre-computed, known-good SHA1 hash of the payload from the .zsync file.
• Compute actual SHA1 hash of the payload file on disk.
• If they match, nothing left to do and program can exit.
• If they don't match, download the full .zsync file.
• (Use the full metadata to see what chunks we have locally, and which need to be downloaded from the remote server.)

At each of those steps where the .zsync file is downloaded, the request is doubled (the bug this issue is about).

And now after #39 has been merged, all four requests are for the full .zsync file.

[TheAssassin]

We should redesign the whole handling of the .zsync file by downloading it once, caching and reusing it from there. The full download should only occur if necessary. We will probably need two variables, one that contains the contents of the file, and one that shows whether it's a partial or complete copy of the .zsync file.