fix(bdd): configure a compute-reachable HTTP invocation callback in the multi-cluster fixture #2245

Problem

The multi-cluster Helmfile BDD can pass initialization and task scenarios, then return HTTP 504 when invoking a function. The worker processes the request but cannot resolve the response callback address across clusters.

The public fixture tests/bdd/fixtures/self-managed-local-bdd-multi.yaml sets gRPC and LLM callback addresses but omits global.workerEndpoints.invocationServiceURL. The self-managed stack already supports that override.

Proposed correction

Configure the HTTP invocation callback through the existing compute-reachable service alias in the multi-cluster fixture. Add a focused regression check and run the full multi-cluster Helmfile BDD. Keep the single-replica limitation of shared callback routing explicit.

Validation

The complete multi-cluster Helmfile BDD passed 10 scenarios and 93 steps with this supported fixture override, including HTTP, Helm-function, gRPC, and LLM invocation. The HTTP failure reproduces without the override. The correction is currently local to the QA worktree and still needs a repository change and focused regression check. This was discovered during #2232 validation after its OpenBao initialization and migration hooks completed successfully.

Relates to #2232.