Fix division by zero (NaN) in gmt_extend_region when +r/+e increment is 0

[Esteban82]

Done with Claude Sonnet 5

Problem

-Rcode+r<inc> and -Rcode+e<inc> allow specifying 0 for one axis to mean "don't adjust that axis" (e.g. +e1/0 to only round/extend in x). However, gmt_extend_region() divided by inc[...] unconditionally when rounding to multiples of the increment, so a 0 increment produced a division by zero → -nan propagated into the resulting region (wesn[YLO]/wesn[YHI]).

Fix

In gmt_extend_region(), skip the floor/ceil rounding for any axis whose increment is 0, leaving that axis's bound untouched instead of dividing by zero.

Tested with

gmt begin test_6434 png
	gmt coast -Bf -W -RIT+r0/2 -Vi
	gmt coast -Bf -W -RIT+e2/0 -Vi -Xw
	gmt coast -Bf -W -RIT+e0/0/0/2 -Vi -Xw
gmt end

The figure is now generated correctly, and the region is reported correctly in the terminal:

coast [INFORMATION]: Region implied by DCW polygons is 6.6149/18.5135/34/48
coast [INFORMATION]: Region implied by DCW polygons is 6/20/35.4922/47.0952
coast [INFORMATION]: Region implied by DCW polygons is 6.6149/18.5135/35.4922/48

Closes #6434