fix: place the guided tour's card from its allocation, and count GTK warnings

The tour measured its card with a height-for-width request whose for_size
came from the natural width minus the margins the tour sets for positioning.
A measure reports margins *included*, so that for_size landed a margin short
of the minimum the box answers with and GTK warned once per frame:

  Trying to measure GtkBox 0x... for width of N, but it needs at least M

(about 9/s for as long as the tour was up). The card now takes its size from
gtk_widget_compute_bounds() once a frame has run -- its allocation rectangle,
not gtk_widget_get_width(), which is only the content box -- and falls back
to a measure with no for_size at all only on a frame with nothing allocated.

GTK is built with structured logging, so its messages reach the log writer
and never the handler API; no existing probe can read one. The exit dump now
counts them (`gtk warnings:`, must read 0) through a writer function chained
to g_log_writer_default(), which leaves stderr unchanged.

docs/claude/verification.md

@@ -24,7 +24,10 @@ touching each `*.gresource.xml` under `src/` and rebuilding; a third `-Wmissing-
 or any other warning is a finding. (The fifth warning that used to appear here was an unused
 `output_built` in `math_probe()`; it was **deleted** as dead code — redundant with the
 `output_after_add` already printed on the same line — which dropped the baseline from five to
-four.) And an exit-dump run emits **no** `Gtk-CRITICAL` at all.
+four.) And an exit-dump run emits **no** `Gtk-CRITICAL` at all, nor any `Gtk-WARNING`: the dump's
+`gtk warnings:` line counts GTK's own log (through a writer function — GTK is built with structured
+logging, so the handler API never sees its messages) and must read `0`; see
+[probe-readings.md](probe-readings.md).
 
 An *incremental* build reads **one** of those three: the generated GResource `.c` is not recompiled
 when only `src/` changed, so its `-Woverlength-strings` does not reappear.