ci: force minikube to use docker runtime to fix buildx 404 errors

[himsngh]

What this PR does / why we need it:

CI pipelines (benchmarks, test-kctrl-gh, upgrade-testing) are failing randomly with a 404 page not found error during the kbld image building phase.

As GitHub Actions rolls out Ubuntu 24.04 runners, Minikube's default internal container runtime has shifted to containerd. When Minikube uses containerd, minikube docker-env relies on a highly experimental SSH proxy bridge. Docker Buildx (which kbld uses to build images) fails to communicate with its isolated moby/buildkit container across this SSH bridge, resulting in the daemon returning a 404 error.

Explicitly append --container-runtime=docker to all minikube start commands across the workflow files. This forces Minikube to use the legacy Docker engine internally, restoring the standard TCP socket communication for docker-env. This allows Buildx to use the native Docker driver without issue, resolving the pipeline flakiness.

Which issue(s) this PR fixes:

Fixes #

Does this PR introduce a user-facing change?

Additional Notes for your reviewer:

This is a tactical fix to immediately unblock the CI pipelines. As a long-term follow-up, we should look into removing the reliance on eval $(minikube docker-env) entirely across our test scripts. Refactoring the tests to build natively and push to a local registry will allow us to safely transition Minikube back to the containerd runtime without breaking Buildx.

Created an issue #1858

Review Checklist:

• Follows the developer guidelines
• Relevant tests are added or updated
• Relevant docs in this repo added or updated
• Relevant carvel.dev docs added or updated in a separate PR and there's
  a link to that PR
• Code is at least as readable and maintainable as it was before this
  change

Additional documentation e.g., Proposal, usage docs, etc.: