Administrators often have to manually identify the source of build failures by looking in each workspace's logs, monitoring dashboards for their provider, or having their users report issues internally. This makes triaging pervasive workspace problems difficult, especially across clouds.

When a build fails, we should include areason for the failure in the control plane logs, simply the returned Terraform error. Right now,I believe we only report theinitialize terraform: exit status 1.

This way admins can filter the coderd logs and set up alerts when a type of error is hit frequently enough.

cc@kylecarbs