Preview the next exercise and count into each set during rests

The run flow's between-exercise rest now previews the next exercise — its
figure and a large "Coming up" name — and the hands-free narration gains a
per-second count-in ("<Exercise>, in 1, 2, 3, GO!") spoken before every set,
plus a "Coming up" announcement as a between-exercise rest begins. A new
Settings > Narration picker chooses the count-in cues, the setup/form read,
or both.

Also fixes spoken cues going silent after the first exercise in a flow split:
the stop-on-teardown moved from the per-exercise view (rebuilt on every
hand-off) to the run host, which stays mounted for the whole run. The audio
session now holds its duck across the per-second count so background music
doesn't pulse between words.

Claude-Session: https://claude.ai/code/session_01BQcEWmAPA78338QuEwRkAh

README.md

@@ -44,14 +44,19 @@ your own iCloud Drive.
   Library's rig data with the working limbs highlighted — the camera slowly
   orbits every exercise while it moves, machine equipment (seats, bars, cables,
   roller pads) has world-space 3D form that turns with the figure, and face-on
-  machines show true hip abduction and torso rotation.
+  machines show true hip abduction and torso rotation. While you rest between
+  exercises in an auto-advancing split, the figure previews the next exercise under
+  a **Coming up** label, so you see what's next before it starts.
 - **Spoken instructions** — tap the speaker in any library exercise to hear its
   full reference read aloud, and turn on **Speak Exercise Cues** in Settings to
-  have the setup and form cues spoken automatically when you start an exercise, so
-  you can follow along hands-free. On-device speech that ducks — never stops — your
-  music. A **Voice** section in Settings picks the voice (preferring any enhanced or
-  premium voices you've installed) and tunes its speed, pitch, and volume, with a
-  live preview.
+  narrate a running workout hands-free. A **Narration** setting chooses the style:
+  *Coming Up & Countdown* announces the next exercise during a rest and counts you
+  into every set (the upcoming exercise, then "in 1, 2, 3, GO!", one word per beat),
+  *Setup & Form Read* speaks each exercise's cues as it starts, or *Both*. On-device
+  speech that ducks — never stops — your music, holding the duck across the count so
+  it doesn't pulse between words. A **Voice** section in Settings picks the voice
+  (preferring any enhanced or premium voices you've installed) and tunes its speed,
+  pitch, and volume, with a live preview.
 - **Machine comfort settings** — machine-based exercises remember your setup
   (seat height, back-rest position, pin position, …), shown on the workout
   screen and editable mid-workout; changes save back to the split for next time.