deps: bump sei-k8s-controller to a6d5bd6; import renderers from test/manifests

[bdchatham]

Summary

sei-protocol/sei-k8s-controller#580 moved the fault catalog and the seiload Job renderer from harness/{faults,bench} to test/manifests/{faults,bench}. The package names and API didn't change. This PR bumps the controller pin to that merge commit and rewrites the imports in chaos/, bench/ and mcp/server.go:

- github.com/sei-protocol/sei-k8s-controller/harness/{faults,bench}
+ github.com/sei-protocol/sei-k8s-controller/test/manifests/{faults,bench}

The pin moves from badf30d to a6d5bd6, which also brings in the controller commits merged in between. go build, go vet and go test ./... all pass. sidecarapi stays on its own replace pin.

Link to Devin session: https://app.devin.ai/sessions/529b10a4cbc7445895aa6189f10b28f7
Open in Devin Desktop: https://app.devin.ai/desktop/session/529b10a4cbc7445895aa6189f10b28f7?variant=devin
Requested by: @bdchatham

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[cursor]

PR Summary

Low Risk
Mechanical dependency bump and import rewrites with unchanged renderer APIs; no security or behavioral changes in seictl itself.

Overview
Tracks sei-k8s-controller after upstream moved the fault catalog and seiload Job renderer from harness/{faults,bench} to test/manifests/{faults,bench} (same package APIs).

The module pin moves from badf30d to a6d5bd6, and seictl import paths are updated in bench/, chaos/, and mcp/server.go so CLI render, chaos, and MCP tools still call the shared templates. No local logic changes beyond those paths and go.sum.

^{Reviewed by Cursor Bugbot for commit 2d833e6. Bugbot is set up for automated code reviews on this repo. Configure here.}

[seidroid]

review found nothing blocking.

[seidroid]

Moves the sei-k8s-controller pin from badf30d to a6d5bd6 and points the faults/bench imports in chaos/, bench/ and mcp/server.go at test/manifests/{faults,bench}. I found nothing blocking: in the upstream compare, controller#580 is a pure rename (the four Go files show +0/-0), nothing in the tree still imports harness/, and the other exported change that reaches this repo is sdk/sei.TaskOutputs gaining a field; I could not build or test here because Go is not installed, and codex's empty reading matches mine, so it added no findings.

Non-blocking

• The bump also brings in 20 other controller commits, which change the SeiNetwork/SeiNode/SeiNodeTask CRD types and schemas. Before merging, run a seictl network apply / seictl node smoke test against a cluster running the matching controller, so presets are checked against the updated CRDs rather than only compiled.

_{seidroid review · decision approve · session 7664f9c0526f462a8209b3987421ed19 · turn resp_claude_8ac995206f665d66a655ec0647a679b8 · item 620f825c5cb15f41a7d531e6d5774550}

Findings: 0 blocking | 1 non-blocking | 0 posted inline