These are relatively minor updates to the Nichols chart (AKA grid) plotting.

It's far from ideal; further possible improvements:

• more tightly fitting the chart to the view (or data) xlim and ylim (the "extent"); the pain here is figuring out which contours to plot, and where to place the labels
  • depending on the limits, some contours will end up with a two visible segments
• recreating the axis on zoom and pan events
  • this might not betoo difficult; Axessupport'xlim_changed' and'ylim_changed' events
• a deluxe version of this would be too add (secondary?) axes, and show the closed-loop gain and phase in the "data value" text (the "x= , y = " on the Matplotlib toolbar)
• I wonder if we could use a customContourSets, and plug into however that labelling is done?

Here are some before-and-after pictures (left is before) of a few example systems.

The new code gives a better result in almost all respects. One visible problem is that the closed-loop phase label for "0°" clashes with the "-40dB" (and similar) closed-loop gain label. The gains become crowded when the gain span is large (final figure), but that is the same in both versions.

I haven't added tests; I'll see how much time I have to do that. Easyish tests:

• check for "tight" phase limits (see first example below)
• check that the axes child text elements are have clip_on (I'd have to not check title & tick labels, presume this is possible)
• check for existence of phase labels, contingent on label_cl_phases flag
• test both branches of theif in_inner_extents, and in fact test the behaviour in general by setting xlim & ylim before calling nichols_grid.
  Please suggest other tests.

Details (commit message)

Clip closed-loop contour labels.

Add labels for constant closed-loop phase contours.

Use smaller of data or view extents when deciding on how big, in
phase, a chart to create.

Use more uniformly spaced closed-loop phase contours, and use more
widely-spaced contours when phase extent is large.

Add optionalax argument, for axes to add grid to.