Fail and surface tar build error when jib:dockerBuild and jib:build

[chanseokoh]

From #626. The execution of jib:dockerBuild becomes stuck.

Finalizing...
Loading to Docker daemon...
<==========---> 83% EXECUTING [1h 19m 22s]

Would be easy to reproduce with jib:dockerBuild by throwing an artificial exception as #626 did originally.

Also, I vaguely remember jib:build did not fail but rather succeeded, which is more concerning.

[TadCordle]

Wait, is the issue here that #626 isn't actually fixed, or just that we need better error feedback if this happens again?

[chanseokoh]

Sorry if it wasn't clear. I expect the underlying issue of #626 (incorrect handling of milti-byte strings) has been fixed. This issue is about not failing but the execution getting stuck forever or appearing to succeed when there was any kind of errors actually building tars.

[TadCordle]

I was able to test this by temporarily reintroducing the problem in #626, and the problem seems to be fixed for the jib:dockerBuild case (probably by #1064). jib:build still succeeds though, but that's because jib:build doesn't ever call the code where the original error occurred (TarStreamBuilder#addByteEntry, formerly TarStreamBuilder#addEntry, which is used by ImageToTarballTranslator, which is only ever used by LoadDockerStep and WriteTarFileStep). So I suspect there's no problem here, especially since we haven't hit this since July.