docs(handover): point at the durable wasm per-call write-up

Author: BuddhiLW

.dev/handover.md

@@ -74,8 +74,9 @@
   Confirmed against zwasm's own `zig build bench-latency` on this host, not
   inferred. Cards: `[CLJW-WASM-ENGINE-DEFAULT]`, `[CLJW-WASM-WORKER-THREAD]`
   (worker threads pay 561 ns instead of 26.8 us, so ~47x may be available
-  cljw-side), `[CLJW-WASM-BENCH-BLIND]`. Probes + numbers:
-  `.dev/bench/ffi_boundary/README.md`.
+  cljw-side, CONFIRMED at 14x), `[CLJW-WASM-BENCH-BLIND]`. **Full write-up incl.
+  the proposed fix and its feasibility check: `.dev/wasm_percall_findings.md`**
+  (durable, written while hive was down). Probes: `.dev/bench/ffi_boundary/`.
 - **`bench/` is stratified and noise-guarded.** shell measures / YAML is the
   datum / Python renders, with `bench/bench_domain.py` owning the vocabulary.
   Harnesses emit `--yaml`; `gen_cross_table.py` (Markdown) and `gen_charts.py`