Don't draw a wait's bar at nothing

A fraction of zero is the absence of a sample, not a measurement of the
work: a bar at nothing for two minutes says the load has not started,
which of one 8 GB into a 12 GB file is false.

Measured raw off the router's event stream today: llama-server reports a
model's load as 0 and then 1 with nothing in between, on both the 0.6B and
the 27B, and reports prompt processing once a batch. So the bar now draws
for a long prompt -- the wait worth watching -- and a load that says
nothing draws none, which is what not knowing should look like.

AGENTS.md

@@ -132,7 +132,12 @@ Module-by-module intent is in PLAN.md's "Backend layout".
   `"return_progress": true` adds to the generation stream. The sample says
   which status it measures so it cannot be drawn under another one, and
   `/loading [seconds] [stages]` / `/reading [seconds]` in an echo session are
-  the rig for the phone's half.
+  the rig for the phone's half. **Zero is never reported**, being the absence
+  of a sample rather than a measurement -- which matters because
+  `llama-server` reports a model's load as 0 and then 1 with nothing between
+  (measured 2026-09-21 on both the 0.6B and the 27B), and reports prompt
+  processing once a batch; so the bar usually draws for a long prompt and not
+  for a load.
   **A turn's wait has two halves and says which** (2026-09-19):
   `SessionStatus::Loading` is the model coming off disk and
   `SessionStatus::Reading` is `llama-server` processing the prompt -- emitted