docs: document the PyPI install path in the guides, not just the README

[FZ2000]

Problem

0.1.0 shipped to PyPI as da-sync and the README was updated to match, but
docs/getting-started.md — the page the README explicitly sends new users to
— still documented only git clone + ./install.sh. The one file written for
people who have never used the tool was the one file that never mentioned the
published package.

Two further gaps surfaced while verifying it:

• install_schedule.sh is in the sdist but not the wheel. Both the
  tutorial and docs/guides/scheduling.md tell you to run
  ./install_schedule.sh, which a pipx user simply does not have. The
  script's own error message pointed only at ./install.sh.
• The ./install.sh transcript was stale. The script prints a
  (N modules) suffix that the documented output had dropped.

Solution

1. Getting-started now offers both install paths, with the clone kept as the
   option that also carries the scheduling script.
2. Both scheduling sections say where to get install_schedule.sh, and note
   it drives whatever da is already on PATH (install_schedule.sh:112
   resolves it with command -v da), so no reinstall is needed. Its error
   message now names both paths.
3. check_doc_references.py asserts any documented module count against
   dacli/, so the number cannot drift again.

Verification

Everything was checked against da-sync 0.1.0 installed from PyPI, run with
XDG_CONFIG_HOME/XDG_STATE_HOME pointed at a scratch directory.

• All 29 README quick-tour commands resolve against the installed binary.
• The docs/commands/config.md transcripts were replayed in the sandbox that
  page documents (/tmp/da-demo, client_id=12345). config show,
  config path, --config, every config get example, the <=8 chars
  masking rule and the exit codes all reproduce byte-for-byte — including
  --unmask writing to stderr, so the documented 2>/dev/null really does
  print nothing.
• docs/commands/sync.md's --help block is byte-identical to the real
  output at the width the docs are generated for (width=78).
• ./install.sh was run with DA_INSTALL_PREFIX redirected, which is how the
  missing (17 modules) suffix was found.
• The new module-count check was watched failing on a wrong number
  (says 99, dacli/ has 17, exit 1) before being kept.
• install_schedule.sh's new error path was executed with da off PATH.
• 868 tests pass; check_doc_flags, check_doc_references,
  check_version_sync, check_discoverability, markdownlint (pinned
  0.45.0 + repo config) and lychee --include-fragments are all clean.

docs/commands/ examples are executed by tests/test_docs_examples.py;
getting-started.md is not, which is how the stale transcript survived.

🤖 Generated with Claude Code