Fix dropped fractional digit in latitude annotations for FORMAT_GEO_MAP with 3 decimals

[Esteban82]

Summary

Fixes #9154.

gmtlib_get_annot_label() (in gmt_support.c) shrinks the leading degree field of the plot format string from 3 digits to 2 for latitudes (max 90°, vs 360° for longitudes) by doing a text substitution: replace every "%3.3d" in the format string with "%2.2d".

The bug: when FORMAT_GEO_MAP requests exactly 3 fractional digits (e.g. ddd.xxxF), the fractional-part printf format built for that case is also literally "%3.3d" — same text, unrelated purpose. The blind, whole-string gmt_strrep() call corrupted that fractional format too, shrinking it from 3 digits to 2. This silently dropped the leading zero of any fraction that happened to be 0** (e.g. .024 became .24), while leaving the value visually plausible (9.020 → 9.20 → rendered as if only two decimals had been requested).

This only manifests for latitudes with exactly 3 fractional digits — ddd.xxF (2 digits) and ddd.xxxxF (4 digits) don't produce the literal string "%3.3d" anywhere else in the format, so they were unaffected. That matches the issue report exactly (3rd panel wrong, 4th panel correct).

Fix

Only replace the leading degree field — which is always at position 0 of the format string — instead of replacing every occurrence of "%3.3d" anywhere in the string.

Test plan

• Reproduced the bug with the PyGMT snippet from the issue, translated to plain GMT (FORMAT_GEO_MAP=ddd.xxxF, -Ba0.0001) — before the fix, y-axis annotations read 9.24°N, 9.23°N, 9.21°N, ... (2 digits, leading zero dropped); after the fix they correctly read 9.024°N, 9.023°N, 9.021°N, ... (3 digits).
• Confirmed ddd.xxF / ddd.xxxxF (2 and 4 decimals) were unaffected before and after.
• Built with ninja gmt (warnings-only, pre-existing and unrelated to this file).
• Ran the doc/scripts ctest suite's -B_geo tests; GMT_-B_geo_2 fails identically on master and on this branch (pre-existing, unrelated font-rendering baseline diff), GMT_-B_geo_1 passes on both.

Assisted-by: Claude Sonnet 5 (medium effort)

[Esteban82]

I tested with this script

R="38/38.0002/9/9.0002"

gmt begin test_9154 png
	gmt set FORMAT_GEO_MAP=ddd.xxF
	gmt basemap -R$R -JM4c -Ba0.0001

	gmt set FORMAT_GEO_MAP=ddd.xxxF
	gmt basemap -R$R -Ba0.0001 -X6c

	gmt set FORMAT_GEO_MAP=ddd.xxxxF
	gmt basemap -R$R -Ba0.0001 -X6c
gmt end

[joa-quim]

We need to revise the walls of text produced by the Agents. That is why I think we should not accept PR submitted by them. The "Test Plan" is clearly not needed here.

[Esteban82]

We need to revise the walls of text produced by the Agents. That is why I think we should not accept PR submitted by them. The "Test Plan" is clearly not needed here.

Yes, I agree.