This adds some synchronization to the maps at the root of the API to avoid duplicated calls to the actual GH REST API. Specifically this is targeted around the two maps, orgs and users. This fix makes the GHPRB jenkins plugin behave much better when there are lots of projects that could build for a specific repo (even if only a few are actually triggered)
There are also a few fixes around GHUser and GHPullRequest
* GHPullRequest was checking a field that may be null (merged_by) when determining whether to fetch details. An unmerged PR would make a bunch of Github API calls for each property accessed.
* Where GHUser was returned in various objects, we weren't going through the caching mechanism at the root, so calls to APIs on GHUSer often resulted in new REST calls. Instead, return from the cache wherever possible.
Setting labels and assignee on Pull requests failed silently, because
the API endpoint being hit contained '/pulls/' rather than '/issues/'.
"Every pull request is an issue, but not every issue is a pull request.
For this reason, “shared” actions for both features, like manipulating
assignees, labels and milestones, are provided within the Issues API."
https://developer.github.com/v3/pulls/#labels-assignees-and-milestones
I was trying to add the flavor of the retrieve method that reads into an existing instance, when I realized that there are just too many orthogonal axes here to rely on overloaded methods.
That calls for a builder pattern, which we already have --- it's called Poster, but it can actually already handle GET and other HTTP requests.
So I'm retiring the retrieveXYZ methods and moving the code into Poster. This is the first step.
Added SmallUser representing reference to user.
Added DetailedPullRequest - when retrieving pull request by id, it has more attributes then pull requests obtained by GHRepository.getPullRequests()
