perf: account for the device memory the renderer holds

Stage 0 of the roadmap asks for a memory baseline before any more
optimisation, and explicitly not from RSS: the allocator and the driver
both retain high-water memory that RSS cannot tell apart from live
resources.

`gpu-live` counts from the sizes this renderer allocated, in the
categories large enough to decide where the next fix goes:

    slots  total  shadow  targets  text_atlas  wire_arena  block_geometry

It is partial by construction and says so, rather than presenting itself
as a total. Hatches, images, meshes and Face3D are not counted yet.

Three of the numbers needed a size that was not reachable before: the
block-geometry chunk records its bytes at build because in storage mode
the buffer belongs to the bind group, the arena exposes its allocation
rather than its occupied prefix — headroom is what a budget has to answer
for — and the text atlas reports its `R8` texel count.

Emitted only when a number moves. `prepare` runs thousands of times in a
session and these change a handful of times, so a per-frame line would
bury the log it is meant to explain.

This answers the question the previous fix left open: whether the arena's
52 MiB per slot, the 16 MiB shadow map every slot allocates whether or not
shadows are on, or the render targets dominate once a paper layout has
four or five viewports on screen.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

src/scene/view/render.rs

@@ -1217,6 +1217,7 @@ impl shader::Primitive for Primitive {
                 self.viewports.len(),
             );
         }
+        report_gpu_live(pipeline);
     }
 
     fn render(
@@ -1568,6 +1569,42 @@ fn solar_direction(
 
 // ── Render-style helpers (impl Scene) ────────────────────────────────────
 
+/// Report what the renderer is holding on the device, when it changes.
+///
+/// Only on change: `prepare` runs thousands of times a session and these
+/// numbers move a handful of times. RSS is deliberately not used — the
+/// allocator and the driver both retain high-water memory that RSS cannot tell
+/// apart from live resources.
+fn report_gpu_live(pipeline: &crate::scene::pipeline::MultiPipeline) {
+    use std::cell::Cell;
+    if !crate::perf::enabled() {
+        return;
+    }
+    thread_local! {
+        static LAST: Cell<Option<crate::scene::pipeline::GpuLiveBytes>> = const { Cell::new(None) };
+    }
+    let now = pipeline.gpu_live_bytes();
+    LAST.with(|last| {
+        if last.get() == Some(now) {
+            return;
+        }
+        last.set(Some(now));
+        const MIB: f64 = 1024.0 * 1024.0;
+        let mib = |bytes: u64| bytes as f64 / MIB;
+        crate::perf_record!(
+            "[perf] gpu-live slots={} total={:.1}MiB shadow={:.1} targets={:.1} \
+text_atlas={:.1} wire_arena={:.1} block_geometry={:.1}",
+            now.slots,
+            mib(now.total()),
+            mib(now.shadow),
+            mib(now.render_targets),
+            mib(now.text_atlas),
+            mib(now.wire_arena),
+            mib(now.block_geometry),
+        );
+    });
+}
+
 impl Scene {
     fn model_tile_vport(&self, index: usize) -> Option<&acadrust::tables::VPort> {
         let rect = self.model_tiles.borrow().get(index)?.rect;