Measurements of resource injection overheads

[ghost]

Feature request

Instrument the dogfooding cluster(s) so that we can start measuring and assessing potential impact of tektoncd/pipeline#3305.

Use case

From the Pipeline issue mentioned above:

Going forward, with the introduction of Tekton Bundles and soon to be other forms of external dependencies, there is a missing concept of the Tekton controllers which is that of "pre-work". That is, reconciliation should not block but reaching out to external services like an image registry is expensive and error prone.

It would be great to start collecting some data to calibrate our understanding of the time and resource costs involved when relying on external services during reconciles. Luckily we've started using bundles in our dogfooding cluster, so we could potentially start collecting that data there.

[ghost]

I'm going to close this one on the basis that the refactoring in tektoncd/pipeline#3305 is moving ahead as part of pipelines' initial remote resolution explorations. We'll figure out how to gather data and draw needed comparisons as part of the experimental controller work rather than tackling this in plumbing I think.