Add flag to automatically attempt to distribute parallel stages across nodes

[abayer]

A problem we hit in Jenkins X's own production builds for the jenkins-x-builders repo is that Kaniko, particularly with very large images and layers, can cause problems with the Kubernetes node's container GC, particularly if there are multiple Kaniko pods running on the node at the same time. This is annoying. =)

Something that might help us with this - and that will hopefully be useful in other contexts - would be to put an affinity policy on the PipelineRun saying "if possible, put the pods for stages in this run on different nodes from each other". It wouldn't be perfect, but it would definitely help.

Now, this is not saying I'm going to add the ability to configure full affinity policies explicitly on the pipeline - first, the only way I can make this work requires using the PipelineRun name (to use in an anti-affinity policy that says "avoid nodes with a pod already on them with the label tekton.dev/pipelineRun=THIS_PIPELINERUN_NAME, and since the name is unique per run, you can't exactly hardcode the name into the jenkins-x.yml. Second, I'm not sure there's enough value in allowing full affinity policy configuration to be worth the added complexity. So I'm just going to add an option (say, distributeParallelAcrossNodes: true) that can be set on a pipeline's options (but not a stage, since we only have control over the affinity policy on the PipelineRun).