v1.151.0

Added

• A Pareto front of four or more objectives is drawn as parallel coordinates.
  plot_pareto_front used to draw a front only for two or three objectives, because
  past three there is no space to put one in. Now each objective gets a vertical axis
  of its own, side by side, and each design on the front is a faint line across them;
  the cursor's design is drawn thick on top, with its value written at each axis.
  Each axis is scaled from its own range, so unlike units share one height, and the
  true values go on as tick labels.

  Every axis is turned so its better end is at the bottom: a maximize axis runs
  larger values downward, a minimize one upward. Lines that run parallel between two
  axes are objectives that agree, and lines that cross are objectives that trade. The
  axis names say (max) or (min), and the title says (better is down) when every
  axis has a direction; an axis whose direction is none or unknown is drawn bare,
  larger values upward. The directions come from the new
  bencher_readout_scatter_directions dataset attribute, which plot_pareto_front
  fills from the new OptimizeResult.target_directions; without it every axis runs
  larger values upward.
  to_rerun_timeline(readout_scatter=...) now takes any number of variables from two up.

• New gallery example example_optimize_pareto_scrub_nd: a quadcopter trading hover
  time, thrust-to-weight, mass, span and noise, with the five-objective front drawn as
  parallel coordinates and the cursor's design shown as a 3-D model.