fix(cli): make session close accept --session and be idempotent

[ankitranjan7]

webcmd session close --session session_abc told you to run webcmd --session session_abc session close — and that command then failed with missing required argument 'session-id'. The suggestion was impossible to satisfy: --session is a root selector everywhere in the CLI except session close, where the id is a positional. A contestant in a recent eval burned three turns on that loop and never closed its session.

What changed

1. session close takes [session-id] (optional) and falls back to the root --session selector. With neither, it errors naming both forms.
2. rejectMisplacedSessionSelectorArgv is command-aware: for session close it rewrites --session <id> into the positional instead of throwing a suggestion that cannot work. Every other command keeps the existing SESSION_SELECTOR_POSITION error.
3. session close is idempotent — an unknown / already-reaped / already-closed Session returns { ok: true, closed: false, alreadyClosed: true, session } and exit 0. InvalidSessionSelectorError (badid) still exits 2; every other failure still propagates.
4. src/daemon/server.ts kept no error code on unhandled command errors, so SESSION_NOT_FOUND arrived at the client as a generic BROWSER_COMMAND with only a message. The code is now preserved on the wire, which is what the idempotency check branches on.
5. --force is untouched.

Before / After

Before (on main):

$ node dist/src/main.js session close --session session_abc
error: SESSION_SELECTOR_POSITION: --session must appear before the command.
       Use: webcmd --session session_abc session close
$ node dist/src/main.js --session session_abc session close
error: missing required argument 'session-id'

After (real output, isolated HOME):

=== webcmd session close --session session_abc
ok: true
closed: false
alreadyClosed: true
session: session_abc
exit=0

=== webcmd --session session_abc session close
ok: true
closed: false
alreadyClosed: true
session: session_abc
exit=0

=== webcmd session close session_00000000-0000-0000-0000-000000000000
ok: true
closed: false
alreadyClosed: true
session: session_00000000-0000-0000-0000-000000000000
exit=0

=== webcmd session close badid
ok: false
error:
  code: INVALID_SESSION_SELECTOR
  message: 'Session selector must be an opaque Session ID: badid'
  help: Run `webcmd session create` and pass the returned `session_...` ID.
  exitCode: 2
exit=2

=== webcmd session close            (no id at all)
ok: false
error:
  code: ARGUMENT
  message: Missing Session ID.
  help: Use `webcmd session close <session-id>` or `webcmd --session <session-id> session close`.
  exitCode: 2
exit=2

Real Session, closed twice, then with --force — all exit 0:

=== close once  → closed: false / alreadyIdle: true   exit=0
=== close twice → closed: false / alreadyIdle: true   exit=0
=== close --force → closed: false / alreadyIdle: true / displaced: null   exit=0

Tests

• src/cli-argv-preprocess.test.ts: --session, --session=, and --profile x prefixed forms all rewrite to session close <id>; a valueless --session is dropped so the command reports both accepted forms.
• src/cli.test.ts: missing Session closes idempotently; root --session selector is accepted; badid still throws INVALID_SESSION_SELECTOR exit 2.
• tests/e2e/management.test.ts: five tests that RUN the CLI — the suggestion round-trip (both spellings, asserting no usage error), unknown id idempotent, real Session closed twice, badid exit 2, and the no-id error text.
• npm run typecheck, npx vitest run --project unit (154 files, 2707 passed), npm run build all green. The new e2e tests pass; management.test.ts > list shows all registered commands fails in this worktree on main too (fixture-plugin install), unrelated to this change.

🤖 Generated with Claude Code

[github-actions]

🟠 Maintainer review suggested — low confidence

The automated review could not reach a fully supported conclusion.

This review is advisory and does not block merging.