ci: add an IPv6-only kind e2e job

[Yuan Gao (ygao-g)]

Do not merge until #958 and #1275 land.

Part of #246.

No CI job exercises IPv6-only, so nothing catches a change that breaks it, and
the IPv6 work is a set of changes that do little apart. This job builds a
single-stack IPv6 kind cluster, installs the full system, and runs the
networking suite, where every test that can tell one address family from
another lives — asserting the cluster really is single-stack before anything is
installed, so a green run cannot be vacuous. It carries no copy of the code it
exercises: dispatch it with a list of pull requests and it merges them onto
main in that order, and the stack lives only for the length of the run.

A fresh IPv6-only cluster resolves nothing from a pod, and a runner has no IPv6
egress of its own, so both the cluster's DNS (#958) and the translator it needs
(#1275) come from hack/; this job asks for them and then reads back what it
got. Setting that up here instead would have meant the job supplying the very
thing it exists to gate, and a green run proving nothing. The cost is that
against main alone it now stops at the NAT64 step rather than reaching the
install — including on the pull_request path that lets this file test itself,
until those two land. #911 has since merged, so the atenet dual-stack bind
needs no gate. The in-cluster registry address still goes stale (#1049). The
job stays out of the e2e-test gate either way, so it never blocks a pull
request.

Demonstrated end to end on #1065, which stacks #958, #753, #1080, #979, #1060,
#1116, #1057 and #1275: the job runs there through pull_request, is green
through all twenty steps, and the vacuity guard reports 11 passed, 0 skipped.

🤖 Generated with Claude Code

[Yuan Gao (ygao-g)]

Dispatch record for this lane. The runs are on the fork; a fork pull request cannot dispatch a workflow upstream.

prs	outcome
958 753 1080 979 1060 1116 1057 1275	11 passed, 0 skipped — 33132033499, and again on a later main in 33187917892
1275	7 passed; TestActorEgress, TestActorEgressHTTPS, TestActorEgressGRPC and TestActorEgressNonStandardPort fail — 33129489452

#1275 on its own brings the cluster up and leaves exactly those four failing, so everything else in the stack is what those four measure. Nothing bisects the eight, so the minimal green set is unmeasured.

The two green runs used a byte-identical e2e-ipv6.yaml on different bases, so the pass is not an artifact of one main.