Documentation / tutorial: File operations as data management

[djdarcy]

Summary

Write a "File operations as data management" tutorial that demonstrates
the v0.3.0 dazzle_filekit.data API in realistic workflows without
ever mentioning the word "primitive".

This is the user-facing counterpart to the v0.2.4 docs/unctools-integration.md
guide, but one layer up: instead of showing how to compose primitives,
it shows how to use the seamless data.* API to do backup, mirror,
sync, and archive operations.

Target audience

Someone who:

• Knows Python but isn't a systems programmer
• Has heard "copy this file from there to here with metadata and a
  hash" and wants one obvious way to do it
• Doesn't want to think about UNC vs drive letter, or whether they're
  on Windows or WSL
• Wants their code to work the same on all three platforms

Structure

1. Intro: the "data management" framing. A file is data with a
   location, not a sequence of bytes at a particular OS path.
2. Hello world: data.copy(src, dst) with a Windows path, a Linux
   path, and a UNC path. Show they all work the same.
3. Rich preservation: data.copy(src, dst, preserve="all") with
   a Windows SDDL example, a Linux xattrs example, and a macOS
   quarantine-skip example.
4. Verification: data.copy(src, dst, verify="sha256") and how the
   hash comparison catches silent corruption.
5. Manifests: data.copy(src, dst, manifest=True) and what the
   sidecar .json file looks like. Show how to restore from the
   manifest later.
6. Mirror a directory: data.sync(src_dir, dst_dir, mode="mirror")
   with a real example (back up ~/projects to an external drive).
7. The object flavor: doc = data.open(path); doc.copy_to(...)
   as an alternative style. Show .metadata, .hash_sha256, .sibling().
8. Graceful degradation: what happens if you don't install preservelib
   or unctools. When do you need them? (Short answer: only for specific
   features; the default behavior works without them.)
9. Cross-platform: show the same tutorial code running identically
   on Windows, Linux, and WSL. Mention is_wsl() as an escape hatch
   for tools that need to know.
10. "I'm coming from shutil": migration guide from shutil.copy*,
    shutil.move, shutil.copytree to the data.* equivalents.

Where it lives

• Primary location: docs/data-management-tutorial.md
• README should link to it from the Quick Start section as
  "For a full walkthrough, see the Data Management tutorial"
• Possibly also published as a blog post or GitHub wiki page (future)

Acceptance criteria

• docs/data-management-tutorial.md exists
• Every code example in the tutorial is runnable -- ideally
  as doctests or as snippets pulled from an actual test file
• The tutorial runs cleanly on both Windows and Linux (the
  cross-platform runner script should cover it)
• README links to the tutorial from the Quick Start section
• The tutorial does NOT use the word "primitive" or explain
  filekit's internal architecture -- it's user-facing only

Dependencies

• Blocked by: #N+1 (data submodule), #N+2 (DataRef API), and
  #N+4 (parse_user_input_path) -- needs the full v0.3.0 API to
  exist before we can document it

Non-goals

• This is NOT the reference documentation -- that's a separate,
  generated-from-docstrings artifact (see API_STABILITY.md for the
  locked surface we'd auto-generate from)
• This is NOT a design document -- no "why we made this decision"
  content, just "here's how to use it"