harness: smooth the busy_tick 255→1 wrap teleport in text_wave comet

[btipling]

Summary

The u8 inv_set_busy_tick phase range 1..255 is incommensurate with the
27-tick / 54-raw comet period in text_wave.zig (SPEED=2, STEPS=3 on N=18).
At the 255→1 wrap (~every 25.6 s at 10 Hz), the localized comet head
jumps ~7 glyphs in a single 100 ms frame — from scalar index 8 back to ~0.7.

The old full-ring cyclicDistance gradient (before #671) hid the same skip
as a phase hitch. The localized sprite makes it visible. The test added in
PR #671 pins the exact values (text_wave.test.zig).

Not a blocker for #671 — just a known artifact of the u8 phase width.

Possible approaches

• Wider phase — use a u16 or u32 tick counter so the comet has more
  steps before wrapping (e.g. wrap at a multiple of N*STEPS).
• Wrap remap — when the host wraps 255→1, remap the incoming phase
  to stay continuous (e.g. phase = wrapCount * 255 + tick).
• Accept the hitch — it is one frame every 25.6 s and may not bother
  operators on tool-using turns that rarely exceed that duration.

Ref

• PR harness: faster, narrower Waiting for model… shimmer (plan #669) #671 review (L1+L9 finding)
• native/harness/src/text_wave.zig headPosition
• native/harness/src/bridge.zig inv_set_busy_tick

[btipling]

Plan: #674 — store the host monotonic u32 (drop the % 255 fold). No protocol bump.

[btipling]

Plan-review landed on the plan issue: #674 — HANDOFF-READY (store the host monotonic u32; drop the % 255 fold; no protocol bump).