docs: scrub stale discover-peers references (PLT-452 follow-up)

[bdchatham]

What

Follow-up to PR-C (#400). PR-C's Coral cross-review (sei-network + kubernetes lenses) flagged docs and one code comment that still present the retired discover-peers task as part of the current bootstrap flow. Peering is wholly controller-owned via the config-apply persistent_peers override — there is no discover-peers task anywhere in the stack.

This PR scrubs the references that misstate current behavior:

File	Change
internal/task/bootstrap_resources.go	bootstrapWaitCommand doc comment no longer lists discover-peers in the bootstrap-task set it waits on (the sei-network finding)
README.md	Full node "Key tasks" chain drops discover-peers
docs/design-seinode-import-volume.md	the "runs exactly as today" task list drops discover-peers
docs/design/composable-genesis.md	both per-node init-plan flows drop the discover-peers step
.agent/runbooks/migrating-validator-to-byo-secrets.md	bootstrap-plan chain + its explanation now credit config-apply (via the controller-resolved peer override) with writing persistent-peers, not discover-peers

Deliberately left (with rationale)

• docs/known-issues-node-alarms.md — a dated incident record. It accurately describes a failure that occurred at discover-peers when that task existed; rewriting it would falsify the historical record.
• Three Go test-comment analogies (common_overrides_test.go:44, controller_test.go:705) and the intentional TestCEL_DiscoverPeers_Removed_Rejected regression guard — internal references that don't misstate current product behavior.

Risk

Docs + one non-functional code comment. No logic change; go build ./... clean.

🤖 Generated with Claude Code

[cursor]

PR Summary

Low Risk
Docs and non-functional comments only; no runtime or bootstrap logic changes.

Overview
Follow-up to retiring discover-peers: peering is set by the controller through the config-apply persistent_peers override, not a sidecar task.

Docs drop discover-peers from bootstrap task chains in README.md, design docs, and the validator migration runbook; the runbook now says config-apply writes base config and folds in peers from the controller-resolved set. Finding 2 in the runbook is reworded to match (peers written on first deploy, not via discover-peers).

Code comments only: bootstrapWaitCommand no longer lists discover-peers among tasks waited on before seid starts; a nodetask test comment no longer compares RestartSeid polling to DiscoverPeers.

known-issues-node-alarms.md keeps the historical discover-peers failure mention and adds a short note that peering is controller-owned today.

^{Reviewed by Cursor Bugbot for commit 3285275. Bugbot is set up for automated code reviews on this repo. Configure here.}