test: DFG Re20 steady rule per review (1 % band, range and slope, 0.5 s samples)

Samples every 0.5 s at synchronized root steps; trailing 5 s window (11
samples); range and |slope*window| each at most 1 % of the official band
width. States the fluid-on-cylinder force sign, reports cd_positive,
defaults to the refined case and uses the review's label.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 71221837f9)

docs/validation/dfg_re20.md

@@ -6,7 +6,7 @@ inflow, initial state, layout and observation are those of the pinned Re100 runn
 (`run_refined_dfg.py`, `re100_helpers.py`, both unchanged); only the lattice
 viscosity is set for Re 20 (root tau 0.56, fine 0.62, uniform 0.62, lattice speed 0.01).
 
-This is a comparison with the official intervals, not a full benchmark
+Label: "compared against the official reference bands", not a full benchmark
 certification: the outlet is pressure-NEQ (`neq_extrap`), not do-nothing, and the
 inlet is local Zou-He.
 
@@ -22,10 +22,13 @@ inlet is local Zou-He.
 ## Bounds and steady rule (frozen before any run)
 
 - End time t* = 60 (30 s physical): 240,000 refined root steps, 480,000 uniform steps.
-- One sample every t* = 0.5; trailing window t* = 10 (5 s, 20 samples).
-- Steady: in the window, max - min of each of Cd, Cl, p_diff is at most one tenth of
-  its official interval width (0.002, 0.00006, 0.00004). Absolute ranges, so the
-  small Cl is never used as a divisor.
+- One sample every t* = 1 (0.5 s) at the end of a synchronized root step; trailing
+  window t* = 10 (last 5 s, 11 samples).
+- Steady: in the window, both max - min and |least-squares slope x window| of each
+  of Cd, Cl, p_diff are at most 1 % of its official interval width (2e-4, 6e-6,
+  4e-6). Absolute limits, so the small Cl is never a divisor. Never relaxed after a run.
+- Force sign: fluid on cylinder, x along the inflow, y up from the bottom wall; Cd > 0
+  and the official Cl > 0 (`cd_positive` is reported).
 - Not steady at t* = 60: status `NOT_STEADY`, no interval verdict. Health limits
   (Mach <= 0.2, rho in [0.5, 1.5]) stop the run with `UNHEALTHY`.
 - Steady: `WITHIN_INTERVALS` or `OUTSIDE_INTERVALS`, with the per-quantity verdict.
@@ -34,10 +37,10 @@ inlet is local Zou-He.
 
 ```
 python tests/validation/run_dfg_re20.py --output <new-dir> --plan
-CUDA_VISIBLE_DEVICES=<n> python tests/validation/run_dfg_re20.py --case both \
+CUDA_VISIBLE_DEVICES=<n> python tests/validation/run_dfg_re20.py --case refined \
     --collision SRT --output <new-dir> --max-wall-seconds <cap>
 ```
 
-`--plan` writes `plan.json` only. A GPU run writes `plan.json` and `result.json`
+Run `--case uniform` only if the refined result is outside a band. `--plan` writes `plan.json` only. A GPU run writes `plan.json` and `result.json`
 (per case: lattice parameters, every sample, window ranges, status, verdict).
 Host contract tests: `tests/unit/test_dfg_re20.py`.