Resource errors should result in a failed deployment with good errors.

Deployment now has top-level error for deployment-level failures

Reintroduce replicatedservice.py as it is used by tests.

Deployment status is now deployment state, which contains both status
and errors, which are set for deployment-level failure.

Some common failures that will show up here are expansion, type
resolution, and reference failures.

Failures during config creation no longer leave deployments undeleteable

Previously a failure would prematurely exit the creation flow without
attaching the manifest to the deployment. This now always attaches the
manifest, and then updates it in place with the outcome of
resourcification.

Resourcifier now processes resources in dependency order based on refs.

For now, references act only as dependencies and do not fill in the
values to which they refer. That will come in a follow-up.

Resources are still processed in a single iteration with no
parallelization.

Fix some document errors

Add support for short github.com types. Refactor manager/manager/types to common/types and reuse them rather than redefine them.

Add support for short github.com types. Refactor manager/manager/types to common/types and reuse them rather than redefine them.

Add support for short github.com types. Refactor manager/manager/types to common/types and reuse them rather than redefine them.

Type resolver now stores resolved imports in the manifest.

Adding manifest support in dm CLI + YAML responses

Omitting parts have the wrong direction

Repository now stores latest manifest in deployment.

Schema updates for Expansion.

Pretty sure you meant from right to left here

Changes to support template moves

Simplify Dockerfiles.

Remove direct access to manifests to prepare for persistence..