Unify the logic of retrying

The following discussion from !54300 (merged) should be addressed:

• @fabiopitino started a discussion: (+2 comments)

  I think the change is ok overall. I'm thinking whether we could take this opportunity to enforce the following invariant: "when a job is added to an existing pipeline we need to ensure previous jobs are marked as retried".

  An idea would be to have a method like below that represents the SSOT for adding new jobs to existing pipeline

  class Ci::Pipeline
    # Use this method to add a job to an existing pipeline
    def add_job!(job)
      raise "Pipeline must be persisted for this method to be used" unless persisted?
  
      # these also ensures other invariants such as
      # a job takes project and ref from the pipeline it belongs to
      job.pipeline = self
      job.project = self.project
      job.ref = self.ref
  
      self.class.transaction do
        job.save!
        job.update_older_statuses_retried!
      end
  
      job
    end
  end

  Then this service could use it as:

  pipeline = build.pipeline
  @status = pipeline.add_job!(GenericCommitStatus.new(
    user: current_user,
    stage: 'deploy',
    name: 'pages:deploy'))

  EDIT: with the guard clause we could also ensure that the job being added is not persisted either.

• !54300 (comment 514109156)

  Ci::RetryBuildService uses the same logic. Should we find a way to remove this duplication?

Availability & Testing