Unique job names for dictobjs

[cisaacstern]

WIP with @jbusecke for ensuring unique job names on dict obj deploys.

There are a few details to work out which I will comment on shortly.

[cisaacstern]

xref pangeo-forge/deploy-recipe-action#17 with changes in action to accommodate this PR

[cisaacstern]

Noting this PR changes the assumption of pangeo-forge/roadmap#49 regarding how dict objects should be declared in a meta.yaml.

We elected to do this because that assumption would prevent a single feedstock from containing multiple recipes representing a mix of dict objects and regular recipe objects, as demonstrated here:

• recipe module: https://github.com/pforgetest/gpcp-from-gcs-feedstock/blob/1ddf63bc5488e8f24d172dafdf69d622842f9589/feedstock/recipe.py#L22-L51
• corresponding meta.yaml: https://github.com/pforgetest/gpcp-from-gcs-feedstock/blob/1ddf63bc5488e8f24d172dafdf69d622842f9589/feedstock/meta.yaml#L7-L10

Generally speaking it's probably not a good idea to just change these assumptions on-the-fly like this, and the roadmap repo linked above is long-since unmaintained, so I think the best path forward here is to first introduce the suggestion discussed starting in pangeo-forge/pangeo-forge-recipes#514 (comment) that https://github.com/pangeo-forge/meta-yaml-schema be consolidated into -runner. Once this is done, the changes to the meta yaml parsing assumptions added in this PR can be paired with a corresponding change to the meta yaml schema, which feels much more intelligible.

[cisaacstern]

Generally speaking it's probably not a good idea to just change these assumptions on-the-fly like this, and the roadmap repo linked above is long-since unmaintained, so I think the best path forward here is to first introduce the suggestion discussed starting in pangeo-forge/pangeo-forge-recipes#514 (comment) that https://github.com/pangeo-forge/meta-yaml-schema be consolidated into -runner. Once this is done, the changes to the meta yaml parsing assumptions added in this PR can be paired with a corresponding change to the meta yaml schema, which feels much more intelligible.

@jbusecke, this is now being tracked in #93, would be great to have you take a look at this, as your current work is reliant on this.

[cisaacstern]

Superseded by #139