Add the rules for auto-cancelling a redundant pipeline

What does this MR do and why?

This MR adds Gitlab::Ci::RedundantPipelines::AutoCancelPolicy. The four callers arrive in later merge requests, so nothing calls this class today.

graph LR
    A["StartPipelineService"] -->|"cancelable_by_newer_pipeline?"| P["AutoCancelPolicy<br/>(this MR)"]
    B["CancelRedundantPipelinesWorker"] -->|"applies?"| P
    C["Ci::Build starts a job"] -->|"protected_after_non_interruptible_job_starts?"| P
    D["PipelineFamily"] -->|"cancel_mode"| P
    P --> E["pipeline metadata,<br/>processing data,<br/>feature flags"]

The class answers four questions about one pipeline. applies? tells you auto-cancel applies to this pipeline on its own. A child pipeline never qualifies, because it can only be cancelled with its parent. cancelable_by_newer_pipeline? tells you a newer pipeline may cancel this one. cancel_mode returns :all, :interruptible, or nil. protected_after_non_interruptible_job_starts? tells you that starting a job that cannot be interrupted puts the pipeline out of reach for good. The MR also adds Ci::Pipeline#interruptible_protected?, which reads the existing p_ci_pipeline_processing_data.interruptible_protected column. The work sits behind the feature flag ci_redundant_pipeline_candidates_cache, type wip, default disabled.

Ci::PipelineCreation::CancelRedundantPipelinesService finds pipelines to cancel by querying the partitioned tables p_ci_pipelines and p_ci_builds on a hot path. This class gathers the rules in one place, so a later MR can answer them without those queries. This is one small piece split out of a proof-of-concept MR.

How to set up and validate locally

bundle exec rspec spec/lib/gitlab/ci/redundant_pipelines/auto_cancel_policy_spec.rb

The suite runs 27 examples and all pass. Nothing else needs validation, because no caller exists.

References

• Discussion issue: #621543
• Parent proof-of-concept MR: !253751

MR acceptance checklist

Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.