Detect *stale* tekton objects prior to a upgrade / reconciliaton

[vdemeester]

A set of stale tekton objects (PipelineRun, …) can block the entire upgrade process (see #2473). The operator should be able to detect if there is stale tekton objects and warn and pause the reconciliation until it gets resolved.

• Stale object are object stuck in terminating state
  • Usually it means there is a deletionTimestamp (in metadata) and a finalizers … and it's old
• To warn user, we should report something in the status.conditions stating that there is some stale objects, with some link or command to detect them.