Runnable support retry

[xyloman]

Description of problem

Some resources such as TaskRun do not expose an ability to perform a retry in the case of the task failing.

Proposed solution

If runnable supported the ability to retry X number of times the execution of a runnable which wraps a TaskRun this would allow for tasks that fail due to networking conditions to be retried.

Given a Runnable wrapping a TaskRun
When the TaskRun fails
Then the runnable should be restamped to retry running the task run again

Example

---
apiVersion: carto.run/v1alpha1
kind: ClusterRunTemplate
metadata:
  name: runtemplate
spec:
  outputs:
    image-ref: .status.taskResults[?(@.name=="myresult")].value
  template:
    apiVersion: tekton.dev/v1beta1
    kind: TaskRun
    metadata:
      generateName: $(runnable.metadata.name)$-
      labels: $(runnable.metadata.labels)$
    spec:
      serviceAccountName: $(runnable.spec.inputs.serviceAccount)$
      retries: 5

Additional Context