A GitHub workflow run with a null head_branch or name leaves a permanent "Pending" row with an empty branch and pins its build at High polling priority #400

What's wrong

The GitHub OpenAPI schema marks workflow-run.head_branch and name as nullable. A run from a fork PR whose fork has since been deleted is the usual case. The GitHub provider maps both without a null guard:

• BuildMonitor/Providers/GitHub.cs:430: gitHubRun.Name.As<RunName>()
• GitHub.cs:520: run.Branch = gitHubRun.HeadBranch.As<BranchName>();

SemanticString.As<T>() throws ArgumentNullException on null (ktsu.Semantics.Strings 5.12.1). The run is also inserted into build.Runs via GetOrCreate(runId, build.CreateRun(...)) (L431) before UpdateRunFromWorkflowAsync fills it in. So when L520 throws, a Run is already stored with Started set and the default RunStatus.Pending. The exception isn't an ApiException, so MakeGitHubRequestAsync doesn't handle it. SyncGuard logs it and the loop moves on.

The Azure DevOps provider already guards the same field (build.SourceBranch ?? string.Empty).

Failure scenario (reproduced in a scratch MSTest)

UpdateRunFromWorkflowAsync is given a WorkflowRun with "head_branch": null, "status": "completed", "conclusion": "success". The call throws ArgumentNullException and leaves run.Status == Pending, IsOngoing == true, Branch == "", build.HasOngoingRun == true and BuildSync.Priority == High.

In the app:

• The table shows a row with an empty branch and a progress ring that never completes. The row is persisted, so it survives restarts.
• Since Poll a build at its running or recent-failure rate whichever branch the run is on [patch] #396, HasOngoingRun keeps the build at High priority. It is polled every 30 s and stays polled in critical-budget mode.
• Every poll throws at the same run, so the other runs in the 10-run page after it are never updated, and an error is logged each time.
• Once the run leaves the page, a RunSync is added for the stuck run. GitHubRuns.Get returns the same null and throws, and because the run still reads as ongoing it is retried every 10 s indefinitely.

Suggested fix / acceptance criteria

• Map HeadBranch ?? string.Empty (or a placeholder such as the head SHA) and Name ?? build.Name.
• Build and fill the Run completely before adding it to build.Runs, so a mapping exception never leaves a half-initialised Pending run behind.
• Test: a WorkflowRun with a null head_branch and/or name maps to a completed run without throwing, and no Pending run is left in build.Runs.