feat(pan): configurable decay / afterglow for the spectrum trace (FR-PAN-14)

[dc0sk]

Stacked on #198 (GPU load) on #197 (TLS) on #196 (POTA).

What

A Spectrum afterglow setting (Settings, milliseconds, 0 = off — the default, otherwise 50–5000). A peak lingers on the trace and fades, so a brief signal (a CW dit, an FT8 tone) can still be seen after it has passed. A ghost trace is drawn under the live one — a faint fill and a dimmer outline.

Design (mine — please overrule)

• Decay is a constant dB per second: an exponential decay of power with time constant τ, so 4.34 dB per τ. Attack is instant; the ghost is never below the live trace.
• Off by default, so nothing changes until you choose.
• Advanced once per row, when the row arrives — deliberately not per drawn frame. perf(pan): redraw at the rate rows arrive, not the display's (FR-PAN-13) #198 makes the window redraw at the row rate, and the look must not depend on how often it redraws; a peak in a row no frame showed isn't lost.
• The ghost restarts when the pan's centre, span or width changes (else it would be drawn at the wrong frequencies), when the setting changes, and when the history is cleared.
• The waterfall is not affected.

Verified

fmt, clippy --all-features, full workspace tests (445, bounded), cargo xtask OK. The maths is tested against closed-form values (4.3429 dB per τ, power ratio 1/e, −50 dB peak after 990 ms is −54.30 ± 0.02), not against the code's own output. 35 mutations; survivors that were real (an "off" state enforced three times over so each masked the others; an untested clamp) are fixed, and one equivalent mutation (a draw-time width check guarding a race no test can reach without a renderer) is commented as such. Two structural guards cover the drawing order and the four hand-offs of the setting (one of which ends in ..Default::default() and would silently reset a forgotten field).

Also fixes the Networks summary saying "of 3 on" after POTA made four.

Not seen on screen — please read

The display was asleep, so how it looks is untested. The fill's opacity (0.10) and the outline's (0.40) are my guesses; that is yours to judge.

Validation parked with DC0SK (not done here)

• Settings → Spectrum afterglow: try 300 and 2000 ms on a live band. Do brief signals linger and fade? Is the fill too faint or too heavy? Does the live trace stay clearly on top?
• Change the pan span / retune: the trail should restart, not smear at the old frequencies.
• Set it back to 0: the trace should look exactly as before.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Ma4bQKqGNFTYyVSEKPcXRk