Bstack is reworking our tc infra code in https://github.com/taskcluster/taskcluster/pull/1005

We need to adopt what we have in cloudops-infra and sops to work with this. Things we know this means so far are

1. Changing the generate_helm task to copy the existing helm from their repo to the right place, instead of running helmit
2. Reorganizing our values files in cloudops-infra and sops to use the changes hierarchy and names for the values

As part of 2 this we may stop relying on deploylib giving us output from terraform and instead manually put that data in our values files.

To validate the results, we can do something like

1. generate_helm against taskcluster before bstack's PR
2. render-helm against that and current cloudops-infra and sops repo
3. save the results out of the way
4. make our changes in branches of cloudops-infra and of sops
5. generate_helm against taskcluster with bstack's PR
6. render-helm against that and the branches of cloudops-infra and sops repo
7. diff the results of that against what we saved in step 3

There shouldn't be any differences that we didn't expect.

Reassigning to edunham since looks like their PR will land after I'm gone.

We decided we will go with manually adding the terraform values.

For the steps where we want to results of rendering old and new templates, it might be a timesaver to delete everything except the templates for the secrets before rendering, since those are what we need to look at to ensure we have all the right values populated.

Unlike our helm, their helm won't fail to render if we miss a value. It will just render an empty string.

Some of this is tracked in edunham's PR at https://github.com/mozilla-services/cloudops-infra/pull/1228