Need ability to retry builds that do not actually start

Sometimes builds fail to start simply due to network reliability in AWS--either the REST API calls to start up the build fail , or, very occasionally, Codebuild fails to pull my container image from ECR (getting 500 internal server errors). These errors usually clear up pretty quickly, and builds can simply be restarted; but since whole pipelines are usually more complicated, it would be nice to be able to specify an option that the plugin could just retry the build for me. The "retry" Jenkins step will not work, simply because I do not want to retry builds that fail legitimately. It seems like the right approach is to make an improvement to this plugin.

Also a side-note, I may actually have bandwidth to work on this myself, if necessary, but if a dedicated maintainer could point me in the right direction initially, I think we could save a lot of time.

Start by tracing the Jenkins plugin's AWS CodeBuild start calls and the REST failures described in the issue, including CodeBuild's ECR image-pull errors. Define an option that retries only builds that fail to start, while leaving legitimate build failures untouched, and verify that the pipeline can complete after transient startup errors.

Written by the indexing model from the issue text.