docs: CLAS PPP below 10 cm without a base station (GIF demo)

[rsasaki0109]

What

Adds a self-contained demo that shows carrier-phase PPP reaching sub-10 cm
positioning without a local base station. Corrections arrive over the QZSS L6
signal (CLAS); the rover only needs its own observations plus broadcast
navigation.

• docs/clas_ppp_no_base_accuracy.gif — animated evidence (scatter + error vs time, 10 cm markers)
• docs/clas_ppp_no_base_accuracy.json — machine-readable metrics
• docs/use_cases/clas_ppp_no_base_station.md — short use-case page (linked in nav)
• scripts/experiments/claslib/generate_clas_ppp_accuracy_gif.py — reproducible generator

Result

Public QZSS-Strategy-Office/claslib 2019-08-27 sample (0627239Q.obs,
sept_2019239.nav, 2019239Q.l6), 3580 one-second epochs, no base station.

Population	Epochs	Median horizontal	P95	Max
FIX	3575	1.37 mm	2.53 mm	10.2 mm
FLOAT	5	60.9 mm	81.3 mm	82.1 mm
All	3580	1.37 mm	2.53 mm	82.1 mm

Every epoch stays below the 10 cm target. Scored against the published antenna
ECEF (-3957235.3717, 3310368.2257, 3737529.7179).

How it is produced

The generator fetches the pinned public sample, runs the reference CLASLIB
rnx2rtkp solver with util/rnx2rtkp/static.conf, parses NMEA GGA, scores it
against the reference, and renders the GIF:

python3 scripts/experiments/claslib/generate_clas_ppp_accuracy_gif.py

It supports --claslib-root (existing checkout) and --nmea (re-render from an
existing solution). mkdocs build --strict passes.

Boundary

• Static, open-sky, Japan-only L6 capture — not a moving/availability result.
• The capability proof uses the public CLASLIB reference solver. The native
  libgnss++ CLAS-OSR path is still tracking this sample and does not yet match
  the reference trajectory; this PR does not claim native parity.
• Capability demonstration, not a survey certification.

Draft pending a look at the framing/naming.