[bmscomp]

Why

Three bugs made the live deploy window unreadable once a deploy had more than a screenful of components:

• A component that had not started showed as "◯ Pending..." with no name.
• The components pane could not scroll, because its content was set only on View's copy of the viewport.
• The pane followed the deploy only while the logs pane had focus, and that pane never has focus at the start. So components below the fold were never seen.

Ctrl-C closed the window and nothing else. The deploy went on with nowhere to print, and the terminal looked hung until it ended.

This is also PR 3 of plans/deploy-enhancement.md, the wizard.

What

The deploy window (deploy_dashboard.go):

• Every row names its component and namespace, pending ones too.
• The pane's content lives in the model. It follows the component being deployed until a key scrolls it.
• One poll at a time. The next tick waits for the poll's answer, where before it ticked every 2 s regardless and stacked kubectl calls up. Only components that have started are polled, and the per-workload pod query is gone.
• A workload's status comes from its ready count: Ready, or Waiting. Before, a healthy one showed its first pod's phase, "Running", in amber.
• Ctrl-C closes the window and says the deploy goes on. Later lines print to the terminal (DashboardController.closed), and a second Ctrl-C ends kates. Stopping the deploy on the first Ctrl-C would kill Helm mid-install, and deploy has no recovery for a release left pending.
• The header prints the percentage once, not "0% 0%".

The wizard:

• Before the first question, it prints the platform, how it was told, the Kubernetes version and the account (detect.Platform). Admin-less runs get this line too, read from what any account may read. The admin-less overlays are unchanged.
• Each component option shows what is installed, in any namespace: · kafka-ui-0.3.2 in kafka.
• The namespaces are asked on two pages at most.
• With TERM=dumb or KATES_ACCESSIBLE=1, every deploy form runs in huh's accessible mode, and the window prints plain lines.
• Component selection says "Monitoring (Prometheus + Grafana)" where it said "Jaeger (Tracing)".

Deliberately not done

• Merging the "Already installed" screen into the component list. The list already shows what is installed, and upgrades are still ticked on that screen.
• Versions on the review's chips: they are listed under the chips.
• A kept release still reads "✔ Ready" in the window. A failed one shows its failureHint only in the final summary.

Tests

New: TestDashboardComponentsPane, TestDashboardPollsOneAtATime, TestDashboardCtrlCClosesWindowOnly, TestInstalledHints, TestPlatform_ReadsOnlyWhatItNeeds.

Run locally:

• in cli: go test -p 1 ./... -race -count=1, go vet ./..., golangci-lint run --new-from-rev=origin/main ./... (0 issues) and govulncheck ./...;
• scripts/check-cli-style.sh and scripts/check-cli-compat.sh;
• in kates: ./mvnw test, which CI runs on CLI changes.

Not run: the wizard and the window in a real TTY against a cluster. The window was checked by rendering it in a test.