perf: instrument the file-open load path behind PERF

The 52 s silent stretch in a large-DWG open could not be attributed:
OpenTimings covered parse / purge / xref / caches, but the FINALIZING
phase -- prepare_open_geometry, which does the first full Model
tessellation and builds the interaction index -- was unmeasured and
reports only two progress ticks, so it was indistinguishable from parse
time.

Add finalize_ms to OpenTimings and a per-stage PERF breakdown:

  open-phases     parse / purge / xref / caches / finalize
  open-finalize   tessellation vs interaction index
  open-index      index build vs prepare_screen
  wires-build     sort cache, visible filter, block cache, classify,
                  memo hits, parallel tessellate, materialize, sort
  resident-post   scene markers, refedit fade, resident layout
  wire-upload     build_wire_buffers, which runs on the render thread

The wires-build line separates the parallel tessellate from the serial
materialize loop that deep-clones every WireModel while the memo retains
a second copy -- the leading hypothesis for both the stall and the
6.45 GB RSS.

No behaviour change. Every timer is gated on crate::perf::enabled(), so
nothing runs without PERF=1 or the in-app PERF panel. The localized
open-summary format string is left alone; changing it would rekey the
string across every locale, so the full breakdown goes to PERF instead.

Co-Authored-By: Claude Opus 5 <[email protected]>

src/io/mod.rs

@@ -383,6 +383,8 @@ async fn open_path_with_phase_attempt(
             purge_ms,
             caches_ms: t_caches.elapsed().as_millis() as u32,
                     xref_ms,
+                    // Filled in after `prepare_open_geometry` below.
+                    finalize_ms: 0,
         };
         caches.corrupt_dropped = dropped;
                 caches.read_stats = Some(read_stats);
@@ -395,11 +397,24 @@ async fn open_path_with_phase_attempt(
                     );
                 }
                 progress2.set(crate::app::OPEN_PHASE_FINALIZING, 9600, 0, 1);
+                let t_finalize = Instant::now();
                 let (prepared_doc, prepared_geometry) =
                     crate::scene::prepare_open_geometry(doc, &caches, model_bg);
+                caches.timings.finalize_ms = t_finalize.elapsed().as_millis() as u32;
                 doc = prepared_doc;
                 caches.prepared_geometry = Some(prepared_geometry);
                 progress2.set(crate::app::OPEN_PHASE_FINALIZING, 9950, 1, 1);
+                // The localized command-line summary predates `finalize_ms` and
+                // is keyed on its format string, so the full breakdown goes out
+                // here instead of changing that key across every locale.
+                crate::perf_record!(
+                    "[perf] open-phases parse={}ms purge={}ms xref={}ms caches={}ms finalize={}ms",
+                    caches.timings.parse_ms,
+                    caches.timings.purge_ms,
+                    caches.timings.xref_ms,
+                    caches.timings.caches_ms,
+                    caches.timings.finalize_ms,
+                );
                     Ok((doc, caches))
                 })()
             }));

src/io/recovery.rs

@@ -373,6 +373,7 @@ impl RecoveryReport {
         lines.push(format!("Validation: {} ms", self.timings.purge_ms));
         lines.push(format!("References: {} ms", self.timings.xref_ms));
         lines.push(format!("Scene caches: {} ms", self.timings.caches_ms));
+        lines.push(format!("Geometry prep: {} ms", self.timings.finalize_ms));
         lines.push(format!("Total: {} ms", self.total_ms));
         if !self.diagnostics.is_empty() {
             lines.push(String::new());