fix(docs): correct os dev default port from 3002 to 3000

[os-zhuang]

Fixes #229

What changed

content/docs/reference/cli.mdx:70 claimed os dev defaults to port 3002. content/docs/quickstart.mdx:200-203 (landed by #143 from a real boot transcript) already says the opposite: os dev asks for 3000, the same port os start uses, and binds the next free port on collision. This PR brings line 70 into agreement with the measured, reviewed value.

-os dev                         # port 3002 by default
-os dev -p 4002
+os dev                         # port 3000 by default
+os dev -p 4000
+```
+
+Note `os dev` asks for the same port `os start` uses (3000). If that port
+is taken it binds the next free one and says so:
+`↪ server bound to port 3001 (requested 3000)`.

The two calls the dispatch asked for

1. How much belongs in the code-block comment. Kept the inline comment terse (# port 3000 by default), matching this file's existing convention (os start's block reads # default port 3000, os serve's reads # port 3000 — no comment in this file carries more than a few words). The collision/rebind behavior — the part that actually surprises someone running os dev next to os start — goes in a short note line under the block instead. That's not a new pattern for this file: os start's block is immediately followed by an unfenced "HOME directory default:" note (lines 40-42) before the next heading, so a short prose line after a code block is already how this file handles "one more fact that doesn't fit in a comment." A two-sentence comment inside the bash fence would have broken the file's terse-comment convention for no reason the reader needs.
2. os dev -p 4002 → 4000. Took it. 4002 read as "3002 plus a bit"; now that the default is corrected, 4000 pairs the same way. Minor note: os serve --port 4000 --no-ui (three lines up) already uses 4000 as its example port, so this introduces a duplicate value in the same file — different flag spelling (--port vs -p) and different command, so I judged it not worth picking a different number just to avoid the repeat, but flagging it since the dispatch asked me to say which way I went and why.

Premise check (dispatch's assumption, tested)

Confirmed 3000-with-next-free-port is still current behavior, checked against @objectstack/cli 17.2.0 (current latest) in /home/user/objectstack, not just the 17.1.0 the quickstart measurement used:

• packages/cli/src/commands/dev.ts: const requestedPort = port ?? '3000'; — default is still 3000.
• packages/cli/src/commands/serve.ts:680: port: Flags.string({ ..., default: readEnvWithDeprecation('OS_PORT', 'PORT', { silent: true }) ?? '3000' }) — same default on the child serve boots.
• The rebind message and PORT_SEARCH_SPAN collision-search logic in serve.ts are unchanged in shape; the most recent CLI commit touching this area (4f9df8f, 17.2.0, #13120) only changed which number the three announcement channels report (the actually-bound port via IHttpServer.getPort(), fixing a --port 0 edge case) — it did not change the default or the collision behavior itself.

No re-arm of #141 needed; this stays the card it was filed as.

Also re-ran the issue's own re-check command on origin/main: cli.mdx was the only English file still carrying the 3002 claim; the seven quickstart.<locale>.mdx siblings still carry it (correctly, per scope — left untouched) and quickstart.mdx itself was already correct (also left untouched).

Scope

Only content/docs/reference/cli.mdx, lines 70-71 plus the new note line. No locale siblings touched, quickstart.mdx untouched, no neighbouring sections tidied.

Verification (all from repo root, per dispatch)

$ pnpm turbo run build --force
...
 Tasks:    1 successful, 1 total
Cached:    0 cached, 1 total

$ pnpm turbo run test --force
...
 Tasks:    1 successful, 1 total
Cached:    0 cached, 1 total

$ node .github/scripts/check-locale-surface.mjs
...
✓ every advertised URL has a source file and every source file is advertised; both `llms` bodies carry every en-only page title and none from the other locales; and no page slug in the content tree contains a dot

All three exit 0 with clean verdict lines (captured before any pipe). Also spot-checked the actual Next.js build output (apps/docs/.next/server/app/llms.mdx/docs/reference/cli.body) contains the corrected text, so this isn't just a source-file edit — it renders. No changeset needed (this repo has no changeset flow / no packages/).

This repo has no skip-changeset label mechanism (that's an objectstack convention) and no packages/, so nothing else to check there.

Verified at commit 58f660a (current HEAD of this branch).

---

Generated by Claude Code