feat(derived data): Allow experimental Pipelines in DebugGroupDerivedDataEndpoint

[kcons]

Rolling out new Features or modified features can be expensive; we need to reprocess the world before we can see them in action.
Unit test coverage is expected for key scenarios, but it's useful to be able to specify non-live Pipelines for processing in non-live contexts to see their impact on known problem groups.
This extends our debug endpoint to allow arbitrary PIpelines to added and requested for evaluation.
By itself, this isn't terribly high impact, but it can be useful for spot-checking fixes against real data, and it starts moving us away from a world of single pipeline assumption. We can apply a similar pattern of pipeline choice to background validation tasks, running them with a potential new pipeline, tagging metrics with the pipeline name, and using comparison of correctness over a significant project or set of groups to inform choices.

Note that in thise case, we're not replacing STATUS, just adding a peer. I plan to follow up to support a "use this new Feature wherever we use STATUS currently" approach which'll be much more interesting; the code was interesting enough to merit independent review.