fix(darcy): correct sign conventions for Darcy flux and velocity

[jcgraciosa]

Summary

Fixes sign-convention inconsistencies in the Darcy and Poisson solvers.
Changes are in solvers.py and constitutive_models.py:

1. SNES_Poisson docstrings — restored missing minus sign: -∇·[κ∇u] = f
2. v_Field — removed double negation; now correctly stores the physical
   Darcy velocity q = -κ(∇h - s) (inflow positive)
3. DarcyFlowModel.flux — now returns physical velocity q = -κ(∇h - s);
   permeability should be set as a positive value (matching literature convention)
4. F1 Template — uses darcy_flux = q consistent with uw3's natural BC
   convention, where add_natural_bc(g, boundary) specifies inward Darcy flux
5. TransientDarcy DFDt — negated to compensate for Fix 3, restoring the
   same net K(∇h−s) that DFDt tracked before any fixes

Validated against the Henry (1964) saltwater intrusion benchmark and the
Badon-Ghijben–Herzberg freshwater lens benchmark — both produce physically
correct results with the correct freshwater inflow direction and wedge geometry.

Tests

• test_1004_DarcyCartesian
• test_1005_TransientDarcyCartesian
• test_0820_template_parameter_propagation

Follow-up

The natural BC sign convention for SNES_TransientDarcy (when f ≠ 0 or
when natural BCs are used on transient problems) will be addressed in a
separate PR.

Closes #214

---

Underworld development team with AI support from Claude Code