Add ai_gateway_multi_default_models feature flag and propagate to AI Gateway

What does this MR do and why?

Implements the gitlab-rails side of the feature flag mechanism to control multi-default_models selection in the AI Gateway (issue #598607 (closed)).

Changes

1. New feature flag (config/feature_flags/ops/ai_gateway_multi_default_models.yml):
   • Type: gitlab_com_derisk
   • Default: false (disabled — safe rollout)
   • Group: group::ai framework
2. Feature flag propagation via x-gitlab-enabled-feature-flags header:
   • ee/lib/api/code_suggestions.rb — code completions and generations
   • ee/lib/api/helpers/duo_workflow_helpers.rb — Duo Workflow requests
   • ee/lib/gitlab/llm/completions/chat.rb — Duo Chat requests
3. Test update (ee/spec/lib/gitlab/llm/completions/chat_spec.rb):
   • Extended the self-managed cloud-connected test to also verify ai_gateway_multi_default_models is not pushed (consistent with expanded_ai_logging behavior)

Rationale

The AI Gateway already has a mechanism to receive feature flags from gitlab-rails via the x-gitlab-enabled-feature-flags header. This MR adds the new flag to all three AI request paths so the AI Gateway can gate multi-model load balancing behind it.

The companion MR in the ai-assist repo reads this flag and uses it to decide whether to call random.choice(default_models) (flag enabled) or always use default_models[0] (flag disabled).

References

• Issue: #598607 (closed)
• Epic: gitlab-org/modelops/applied-ml/code-suggestions&39
• Rollout Issue: #601113

Screenshots or screen recordings

N/A — backend-only change

How to set up and validate locally

1. Enable the feature flag in rails console:

   Feature.enable(:ai_gateway_multi_default_models)

2. Make a code suggestion or Duo Chat request and verify the x-gitlab-enabled-feature-flags header sent to AI Gateway includes ai_gateway_multi_default_models.

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.