Make the deploy resilient to a bad API token

[kwsantiago]

The deploy on #6's merge failed with HTTP 401 because the API token was invalid, and list_open_prs was uncaught, so one bad data source took down the entire site build.

• Build the open-Knots-PRs view first, so its ~150 calls get a fresh rate-limit budget before the large spec ingest (which can exhaust a token's hourly limit).
• Wrap it so any failure (bad token, rate limit, network) logs a warning and skips just that view; the spec snapshot still deploys.

Verified locally: a bogus token now logs skipping open-Knots-PRs view and still writes a working index.html; a good token builds with the Knots view first.