web,infra: two loose ends from #962 — `url_configured` has no consumer, and the chart guard's comment names the wrong helper

[baakhoff]

Two small things the merge review of #962 (PRs #964/#963) found and deliberately left alone —
neither is worth a behaviour change on its own, both are worth correcting when the files are next
open.

1. url_configured is parsed by the shell and read by nothing

GET /platform/v1/llm/local-runtime returns {state, url_configured}. The web parses
url_configured into LocalRuntimeStatus (services/web/src/lib/contracts.ts:41) and then no
component ever reads it — every decision is made from state alone.

It is harmless because it is derivable: url_configured is false exactly when state is
absent. But that is also the argument that it carries no information for the client, and a
field on a young contract with no consumer either grows one or should be understood as
core-side diagnostics rather than part of the client contract. Worth a decision either way
before anything else starts depending on it.

2. The chart guard's comment names the wrong helper

infra/k8s/epicurus/templates/_helpers.tpl:169 says the hosted-alias guard mirrors
providers.is_hosted. It actually mirrors providers.resolve, and resolve is the correct
thing to mirror — the question the guard asks is "does this model id route to the local
runtime", which is exactly what resolve answers.

The two differ on a malformed id: claude/ (a known prefix with an empty model part) passes the
chart guard, while is_hosted("claude/") is False. So the code is right and only the comment
is wrong — but the comment is what the next person will trust when they change the list.

Comment-only fix; no chart version bump needed if it rides along with another chart change.