chassis: Use pinv instead of solving least squares

[auscompgeek]

Given:

v_wheels = [v_1x v_1y v_2x v_2y v_3x v_3y v_4x v_4y].T
v_robot = [v_x v_y w].T
    [1 0 -r_1y]
    [0 1  r_1x]
    [1 0 -r_2y]
A = [0 1  r_2x]
    [1 0 -r_3y]
    [0 1  r_3x]
    [1 0 -r_4y]
    [0 1  r_4x]

v_wheels = A v_robot

where:

• v_n denotes the Cartesian velocity of module n, and
• r_n denotes the Cartesian centre of rotation of module n,

we can left-multiply the pseudo-inverse of A to solve for v_robot.

I presume this should be significantly faster than solving least squares on the fly.

This is currently on top of #75 (otherwise the tests will fail).

[codecov-io]

Codecov Report

Merging #76 into master will decrease coverage by 0.02%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master      #76      +/-   ##
==========================================
- Coverage   83.68%   83.66%   -0.03%     
==========================================
  Files          21       21              
  Lines        1606     1604       -2     
  Branches      160      160              
==========================================
- Hits         1344     1342       -2     
  Misses        220      220              
  Partials       42       42

Impacted Files	Coverage Δ
pyswervedrive/chassis.py	100% <100%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update dda4b60...32b6a36. Read the comment docs.