Add experimental kinematic flow coupling

[marpaia]

Purpose

This PR depends on the native flow fields introduced by #3 and defines how those fields move rod-shaped cells and how models replace spatial reaction fields during a run. It is the first stack layer that couples solved flow to cell geometry.

Coupling method

• Endpoint-sampled kinematic drift: each non-fixed rod samples the face-staggered velocity field at both capsule centerline endpoints. Translation is the mean endpoint velocity times dt; rotation is dt * (axis cross (v2 - v1)) / length, capped by the mechanics rotation limit. Degenerate rods do not rotate.
• Operator order: drift runs after growth and before contact relaxation. Contact relaxation then resolves cell-cell and cell-wall overlaps introduced by the drift step. Sampling clamps to the grid boundary, and an all-solid stencil returns zero velocity.
• Backend boundary: Add native CPU, Metal, and CUDA microfluidic flow solvers #3 computes the velocity field through the selected CPU, Metal, or CUDA Simulation. Drift consumes the committed field through the common simulation contract, so the kinematic rule is identical across backends.
• Checkpoint behavior: flow_drift is validated and restored with the mechanics configuration. A regression test compares uninterrupted motion with save, load, and resume.
• Reaction feedback: Simulation.set_signal_reaction validates and atomically replaces affine grid source and loss fields, allowing a controller to update colony-dependent sinks during a run.

Scientific scope

• Drift is passive kinematic advection, not force-balanced fluid-structure interaction. It contains no hydrodynamic drag law, wall-dependent slip, fluid force or torque in the mechanics solve, hydrodynamic cell-cell interaction, or Brownian motion.
• The split update is first order in time. Quantitative studies must keep max_speed * dt small relative to a cell radius and report timestep sensitivity.
• Colony-to-flow feedback retains the phenomenological mobility or drag closures from Add native CPU, Metal, and CUDA microfluidic flow solvers #3 and requires coefficient, grid, and refresh-cadence sensitivity analysis.

Stack

PR 3 of 5.

Base: #3 (marpaia/microfluidics-flow-solvers).

Next: #5 (marpaia/microfluidics-device-assets).

Validation at 47f4c33

• Python suite: 265 passed, 42 skipped
• Full pre-commit suite: passed
• The inherited CPU and Metal native flow and conformance gates passed after the rebase
• CUDA 12.8 compile and link gate passed; no NVIDIA runtime was exercised