Correct what a docs walkthrough measured, and show the fabric's state

[mbakalarski]

Five things found by following docs/ step by step on a live lab rather than
by reading the pages.

• lab.md printed an image ID nobody can match. docker import builds a
  fresh layer every time, so the same cEOS tarball produces a different ID on
  every machine and on every re-import. It is a placeholder now; the
  repository, tag and size are what a reader should compare.

• The scenarios overview applied one scenario and removed another. The
  apply block used router, the teardown block used jsonrpc, so following
  both in order deletes resources that were never created:

  Error from server (NotFound): ... bgpglobals.eos.netclab.dev "bgpglobal-on-ceos-jsonrpc" not found
  

  Both blocks now use one scenario, and the page says outright that the
  directory name is the only thing that changes.

• router.md said "the scenarios below this one", which the navigation
  puts above it. The contrast is about what Router does and does not need
  them named.

• fabric.md said the whole scenario is one patch. The patch is what this
  repository adds; the scenario is also the render and the push, and the page
  showed neither. It now walks the Fabric and the eight Devices, the
  deviceCount and pyavd's verdict, and status.push on a device that was
  actually configured.

• fabric.md quoted a device without a command. Every other page gives the
  command whose output it prints. This one showed the leaf's configuration with
  no way to reproduce it; show running-config is where those lines come from,
  scattered through it rather than contiguous.

Everything added here was measured against a running lab today, not carried
over from an earlier session. mkdocs build --strict passes.

Not addressed here, because it is an upstream change: kubectl -n avd get netclab lists nothing, since function-avd's Fabric and Device XRDs carry no
categories. The explicit kinds used above stay correct either way.

🤖 Generated with Claude Code