chore: bump controller image to e99c04c (replace-pod for chain-upgrade rollouts)

[bdchatham]

Summary

• Pulls in feat(planner): add replace-pod task to NodeUpdate plan (closes #211) #216 — NodeUpdate plans now include a replace-pod task that proactively deletes pods at the old StatefulSet revision.
• StatefulSets generated by this controller now use OnDelete update strategy (pod lifecycle is the SeiNode controller's responsibility, not the StatefulSet controller's).
• Unblocks the chain-upgrade rollout pattern where seid is intentionally unready (halted at upgrade height) and native RollingUpdate refuses to disturb the unready pod.

Closes #211 once this lands and rolls out.

Rollback semantics

OnDelete is set in the StatefulSet spec (internal/noderesource/), so reverting the controller container alone won't flip the strategy back — existing StatefulSets stay on OnDelete. To fully revert, also delete and recreate the affected SeiNode resources, or hand-patch the StatefulSet spec.

Test plan

• Confirm Flux reconciles the new image on dev / harbor / prod
• Watch the next NodeUpdate plan (image bump on a Running node) — expect plan steps apply-statefulset → apply-service → replace-pod → observe-image → mark-ready
• Verify pod at old controller-revision-hash is deleted by replace-pod and StatefulSet recreates it at the new revision
• Confirm steady-state behavior: no plan built when spec.image == status.currentImage

---

Note

Low Risk
Low risk manifest-only change, but it will roll out a new controller binary; any behavior change depends entirely on the new image contents.

Overview
Updates the controller-manager Deployment (config/manager/manager.yaml) to use a new sei-k8s-controller image tag (069a866… → e99c04c…).

^{Reviewed by Cursor Bugbot for commit d4d1e52. Bugbot is set up for automated code reviews on this repo. Configure here.}