Fix thin polar sectors that end up parallel to an axis (#7059)

[joa-quim]

A single polar axis is a sector so thin that its bounding box has a width but no height. Rotating it by 90 degrees swaps the two and broke both the scale and the annotations:

• gmtmap_setinfo matched -JP to the map width, which for a sector standing on end is nothing but round-off, so -JP12c+t90 produced a plot 12 x 3437746 cm. It now checks the dimension it was asked to match and falls back to the other one when the first is under 0.01 % of it, with a warning (and a pointer to -Jp for polar plots). A fully degenerate map no longer divides by zero.
• n_lon_nodes/n_lat_nodes are derived from the map dimensions and rounded down to zero for a map with no width, so the boundary-walking loops in gmtlib_map_latcross/gmtlib_map_loncross never ran and no ticks or annotations were placed. The three places that derive them now floor them at 2.

Adds test/psbasemap/polar_thin_sector.sh and documents the meaning of the -JP size for sectors, plus the -Jp recipe for radar plots.

Fix #7059

Supersedes #9131

Assisted-by: GPT Terra + Claude Opus 5.0

[Esteban82]

I tested it and the test works fine.

But I run the command of #7059 and I got this figure

I suspect that the origin of the second axis should be at 0.0.

[joa-quim]

Damn it (thanks for testing).

[Esteban82]

I tested further. With -45 works fine:

gmt begin radar png
	gmt basemap -R-0.0001/0.0001/0/1 -Jp12c -Bya
	gmt basemap -Jp12c+t-45 -Bya
	gmt basemap -Jp12c+t-90 -Bya
gmt end

---

But fails with +t45.

gmt begin radar png
	gmt basemap -R-0.0001/0.0001/0/1 -Jp12c -Bya
	gmt basemap -Jp12c+t45 -Bya
	gmt basemap -Jp12c+t-90 -Bya
gmt end

[joa-quim]

It smells like the thing has tailored the code to comply with the example given and not with a general solution.