engine: downstream MTU and downstream train in the measurement document

Three facts the client cannot produce alone, kept deliberately separate:
mtu.pmtud_down (largest datagram that arrives unfragmented — meaningful only
because the server sets DF), mtu.frag_delivery (whether larger ones arrive once
fragmentation is allowed), and train.udp_downstream (loss, reordering and
arrival spacing in the download direction, which a round trip cannot separate
from upstream loss).

ServerMeasurement now runs them on the same ProbeSession as the echo train. It
had to: a fresh session restarts client-side sequence numbers and the server's
anti-replay window discards the lot, so the re-primed source is never recorded
and every granted send goes to a socket that has already closed. That produced
four confidently-wrong FAILED tests and a RED verdict on a healthy network.

Live against fmr: path MTU 1500, fragments to 4000, 100/100 downstream, GREEN.

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

web/public/assets/wordmark-on-dark.svg

@@ -0,0 +1,17 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="-10 -92 433 104">
+  <!-- "echolot" — hand-drawn monoline letterforms (no font dependency). For dark grounds. -->
+  <g fill="none" stroke="#E8F4F2" stroke-width="11" stroke-linecap="round">
+    <path d="M 0 -24 H 48"/>
+    <path d="M 48 -24 A 24 24 0 1 0 40.97 -7.03"/>
+    <path d="M 110.97 -40.97 A 24 24 0 1 0 110.97 -7.03"/>
+    <path d="M 140 -76 V 0"/>
+    <path d="M 140 -24 A 24 24 0 0 1 188 -24 L 188 0"/>
+    <circle cx="234" cy="-24" r="24"/>
+    <path d="M 280 -76 V 0"/>
+    <circle cx="326" cy="-24" r="24" stroke="#35E0C4"/>
+    <path d="M 372 -48 H 402"/>
+    <path d="M 387 -68 V 0"/>
+  </g>
+  <!-- the finding -->
+  <circle cx="326" cy="-24" r="6.5" fill="#FFB454"/>
+</svg>