Add fix_pipeline_experimental feature flag

Related to Experiment: Run Fix Pipeline as a single autono... (#604562)

What

Adds the fix_pipeline_experimental experiment feature flag, which selects the experimental single-agent Fix Pipeline flow on the AI gateway (fix_pipeline/experimental) instead of the default/next flow.

Precedence: fix_pipeline_experimental takes precedence over fix_pipeline_next — when both are enabled, the experimental flow wins.

Changes

• ee/config/feature_flags/experiment/fix_pipeline_experimental.yml — new flag, mirroring fix_pipeline_next (type: experiment, group::pipeline execution, default_enabled: false).
• ee/lib/ai/duo_workflows/foundational_flow_start_params_resolver.rb — gates the experimental flow ahead of the next branch, using the same container / container.root_ancestor scope. Returns fix_pipeline/experimental (the flow lives under agent_platform/experimental with flow_config_id: fix_pipeline, so the schema version is experimental, not v1).
• Specs — foundational_flow_start_params_resolver_spec.rb covers project scope, root-ancestor scope, and explicit precedence over next.

Dependency

The gateway-side flow it targets is added/renamed to fix_pipeline/experimental in ai-assist!5921. This MR is safe to merge independently because the flag defaults to off — the experimental path is only taken once the flag is enabled, which should wait until !5921 (merged) is deployed.

Testing

• Specs added/updated for the precedence logic.
• Pre-push hooks (rubocop, danger, secrets-detection, openapi_docs, etc.) pass locally.

🤖 Generated with Claude Code