Deploy MSI build artifacts

[Boddlnagg]

This enables deployment of rustup.msi and the corresponding SHA of the embedded executable. I don't know exactly how the corresponding AppVeyor settings work, so whoever is responsible for the original setup should have a look at this, since I'm not able to test it locally.

[brson]

The windows failure looks legit. Not sure what's going on with the unix builds, but it looks similar to something I saw building an older cargo rev recently. It looks like the hash of the openssl tarball changed!

I suspect the appveyor settings for 'artifacts' and 'deployment' aren't quite right but I'll have to experiment with it myself.

Thanks @Boddlnagg.

[Boddlnagg]

I'm not sure if this will actually work, since the artifacts rustup-msi and rustup-msi-exe-sha will only exist for the BUILD_MSI configuration. But maybe AppVeyor does the right thing and just ignores the artifacts if the files in dist don't exist (it looks like that's what's happening for the cases where prepare-deploy-appveyor.ps1 just exits without putting anything in dist).

[brson]

The appveyor failure is unrelated to this pr. Looks like a race in a pre-existing test. Trying to figure it out now.

[brson]

I've also fixed the travis failures in a separate PR.

[Boddlnagg]

I just rebased this and removed some unrelated changes that belong to another PR. I have more changes ready, but I think I'm going to wait until this is merged.

[brson]

Thanks @Boddlnagg. Fighting fires now but haven't forgotten.

[bors]

☔ The latest upstream changes (presumably #776) made this pull request unmergeable. Please resolve the merge conflicts.

[Diggsey]

@brson We need to do something with this - can I do anything to help?

[brson]

@Diggsey Yeah we do.

I'm afraid I haven't looked too closely at the solution here, so I'm not sure the state of it.

If you wanted you could try to get it building again and seeing if you can get it to pass the CI.

[Boddlnagg]

I can offer some help if needed, though my time is limited.

I have more changes ready, but I think I'm going to wait until this is merged.

According to this earlier comment of mine, I also have some more changes that build upon this PR ... I am going to see what that was and can try to prepare a PR.

[Boddlnagg]

@brson On another note, I think the important thing with respect to MSI build and Windows support is that someone needs to decide about what it should look like from a more general perspective. I have outlined some thoughts here.

(In short: My idea and proposal is to have an MSI-based installer replace rustup-init.exe and power rustup self update, but not more than that. IDEs and editor plugins should provide GUIs for the toolchain management tasks themselves by wrapping rustup.)

[brson]

@Boddlnagg thanks for the updates. Sorry for avoiding this for so long. I left a comment on the msi issue in response.

[alexcrichton]

Closing in favor of #1211 which looks like it's got some more action