Remove the snapshot command and enhance reader to try Firefox Reader
View first, falling back to raw Turndown conversion of document.body
when Reader View fails or is skipped via --no-reader.
- reader always returns markdown by default (--format=json for structured)
- JSON output includes method: 'reader' | 'raw' to signal extraction path
- --no-reader skips Reader View (stays on loaded page, preserving JS mutations)
- Add @ts-nocheck to test/smoke.js and exclude test/ from tsconfig
- Update all tests from snapshot to reader with --no-reader for data URIs
- Update AGENTS.md and help text
BREAKING CHANGE: snapshot subcommand removed; use reader instead.
The search token validation test only cleared KAGI_TOKEN from the
environment but still loaded the user config file, which could supply
the token and cause the test to pass incorrectly. Pass
--config=/nonexistent/path so loadConfig returns an empty object.
Also includes search command improvements: markdown/json format output
and --format flag.
