Fix pssac drawing a wrong short line (#5954)

[Esteban82]

Done with Claude (Sonnet 5, high reasoning effort).

Summary

Fixes #5954, where pssac/sac occasionally draws a short, disconnected
stray line segment near a trace peak.

Two bugs combined to produce this:

• pssac.c never wrapped its line/fill drawing in gmt_map_clip_on/ gmt_map_clip_off, unlike other line-plotting modules (psxy, pswiggle, ...), so traces briefly exceeding the plotted y-range were never clipped to the frame.
• The shared gmt_plot_line pen-width border-extension helper (gmtplot_get_outside_point_extension in gmt_plot.c) computes an extension length via W / tan(angle). When a line crosses the border at a near-tangent angle — which happens whenever a smooth waveform peak just barely pokes above/below the plotted range, as in the reporter's SAC trace — this blows up to a length far beyond the segment that defines the crossing direction, producing a visible stray line. This is a generic bug (reproduced identically via plain psxy on the same sample values), not specific to SAC data.

Fix

• gmt_plot.c: cap the extension length to the length of the segment that defines the crossing direction (L = MIN(L, hypot(dx, dy))).
• seis/pssac.c: add the missing gmt_map_clip_on/gmt_map_clip_off around the per-trace plotting loop.

Checked visually with:

gmt begin plot_test_5954 png
    gmt plot test.txt -JX10c/5c -W1p -R-8.22653e-14/3.935039/9.907071e-08/0.43307
gmt end

gmt begin sac_test_5954 png
    gmt sac -JX10c/1.1c -R-2.23517e-07/20/-4.80721e-06/4.30574e-06 bug.sac
gmt end

The files are in #5954.