feat(gds-control): backward reachable set computation

[rororowyourboat]

Problem

gds-control models control system structure (states, inputs, sensors, controllers) but does not compute reachable sets. Backward reachability is a key analysis tool for safety verification and differential games.

Proposed Solution

Add reachability analysis to gds-control:

• Backward reachable set computation from target sets
• Isochrone extraction (level sets of the value function)
• Compatible with continuous-time models (feat(gds-sim): continuous-time ODE integration support #122)
• Visualization support (feat(gds-viz): dynamical systems visualization (phase portraits, trajectories) #126)

Motivation

The Homicidal Chauffeur notebook computes backward reachable sets by integrating 80 dense trajectories backward from the capture circle, then extracting isochrones at T=1,2,4,6,8,12.

Dependencies

• feat(gds-sim): continuous-time ODE integration support #122 (continuous-time ODE integration — backward integration)
• feat(gds-viz): dynamical systems visualization (phase portraits, trajectories) #126 (dynamical visualization — isochrone plots)

[rororowyourboat]

Completed in b0ba22e. Backward reachable set computation in gds-analysis:

• backward_reachable_set(): backward ODE integration from target set boundary
• extract_isochrones(): level set extraction at specified times
• 9 tests including HC capture circle and energy conservation checks
• gds-continuous[scipy] + numpy as optional extras