Move package retrieval code into new IUpdateSource abstraction

[caesay]

In UpdateManager, there is currently added complexity around checking for, and downloading updates as there are two codepaths that can be followed. One for packages on the file system and one for packages at a remote URL.

Additionally, the GitHub implementation tries to hack an implementation into the "remote URL" path, even though this is not optimal, as it's not possible for us to customize how UpdateManager performs the actual web request for the package (such as adding a header). This means private repositories are not supported, as we can't just follow the browser url's to download files. (https://github.com/clowd/Clowd.Squirrel/issues/50)

This pull request creates a new abstraction, IUpdateSource containing the following two methods: GetReleaseFeed and DownloadReleaseEntry.

All of the code within UpdateManager to handle the specifics of retrieving packages has been removed, resulting in a dramatic simplification of that class. There are now a handful of package sources, SimpleFileSource, SimpleWebSource, and GithubSource. It is also extremely easy for a third party developer to create a new package source and customize every detail about how metadata and packages are downloaded.

GithubUpdateManager is no longer needed, but it's not harming anything so I have left it for backwards compatibility. For now it's just a short-cut to perform the following:

new UpdateManager(new GithubSource("repoUrl", "accessToken", prerelease));

[Mat-Stevenson-Simworx]

I have tested the changes for my use case and they seem to be working as expected. Changes looks good, I like the IUpdateSource. I was wondering whether IFileDownloader serves much perpose anymore and it might be good to depricate it from the public api. Just seems like an added complexity now. For example if i was to implement another source that used a web api i feel i might have to by pass it and just juse a http client directly to get access to more functionallity.

[caesay]

I am glad that it is working for you! I also considered removing IFileDownloader like you suggest, however I believe it can still be useful:

• If you need to add a web proxy (eg. to escape a private intranet), you can simply derive from HttpClientFileDownloader, add the proxy configuration, and pass it into a SimpleWebSource constructor.
• Other parts of Squirrel still need to download simple files and IUpdateSource is not suitable for this. For example - when downloading the .net runtime installers.
• SimpleWebSource and GithubSource still need a capable http client, and FileDownloader does a good job of wrapping up common functionality into just a few simple methods so it's logic does not need to be duplicated in each source.

So the approach to extending Squirrel in general should be:

• If you just need to add headers to the request, or add proxy configuration, you can derive from IFileDownloader or one of the concrete classes.
• If you need more customization - eg. retrieving package metadata from a custom API or device, you can derive from IUpdateSource

If you are writing a new IUpdateSource, you do not need to use the FileDownloader for anything, but it still serves a purpose for providing an easy way to extend built-in functionality.