Add rotational viscous damping fix (torque = -gamma * omega)

[SueHeir]

Summary

• Adds RotationalViscousDef and apply_rotational_viscous system to mddem_fixes, the rotational analog of the existing translational viscous damping
• Applies damping torque proportional to angular velocity: τ = -γ * ω, removing angular kinetic energy to help particles stop spinning during settling
• Registered at ScheduleSet::PostForce (same as translational viscous); only active when [[rotational_viscous]] config is present (zero overhead otherwise)
• Accesses DemAtom omega/torque via AtomDataRegistry

Configuration

[[rotational_viscous]]
group = "all"
gamma = 0.001  # angular damping coefficient

Test plan

• Damping torque opposes angular velocity direction
• Zero angular velocity produces zero damping torque
• Group filtering correctly excludes non-group atoms
• Torque magnitude scales linearly with gamma and omega
• All 12 tests pass (cargo test -p mddem_fixes --no-default-features)

🤖 Generated with Claude Code

[SueHeir]

Maintainer Review: ❌ CHANGES REQUESTED

Good physics implementation, but there's an architectural layering issue.

Issue: mddem_fixes → dem_atom dependency

mddem_fixes is an mddem_* (shared infrastructure) crate. This PR adds a dependency on dem_atom (a dem_* domain crate). This violates the project's dependency layering:

• mddem_* crates should work for both MD and DEM users
• An MD-only project using mddem_fixes (for addforce, setforce, viscous, gravity) would now pull in dem_atom unnecessarily

Files: crates/mddem_fixes/Cargo.toml (dem_atom dependency added), crates/mddem_fixes/src/lib.rs (DemAtom imported)

Suggested fix

Move RotationalViscousDef and apply_rotational_viscous to a DEM-specific location:

Option A (preferred): Create a small dem_fixes crate for DEM-specific fixes, keeping mddem_fixes clean for shared infrastructure.

Option B: Use a Cargo feature flag in mddem_fixes:

[features]
dem = ["dem_atom"]

Gate the rotational viscous code behind #[cfg(feature = "dem")].

What's good

• Physics is correct: τ = -γ * ω is standard rotational viscous damping
• Config pattern ([[rotational_viscous]]) follows conventions
• Tests are well-written with analytical validation
• AtomDataRegistry access pattern is correct

The fix itself is clean — it just needs to live in the right place.

[SueHeir]

Fix: Moved rotational viscous from mddem_fixes to dem_granular

Addressed the architectural layering issue:

• Removed dem_atom dependency from mddem_fixes (stays mddem_* clean)
• Created dem_granular::rotational_viscous module with its own RotationalViscousPlugin
• Re-exported from mddem::prelude
• All 4 tests passing, full workspace builds clean