UI/rf waterfall sdr style

[vnxdtzip]

Proposed Changes

This reworks RF Waterfall onto the shared SpectrumPlot component and adds an
opt-in SDR colourmap to that component, so the screen reads like desktop SDR
software while its chrome stays consistent with the rest of Bruce.

Two parts:

1. SpectrumPlot gains an optional SDR waterfall colourmap:

   bool begin(const String &title, bool sdrWaterfall = false);
   

   When true, the waterfall ramp is the classic SDR gradient (noise floor
   blended into the panel, then blue, cyan, green, yellow, red, up to a
   white-hot peak) instead of the theme heat ramp. The trace on top stays
   theme-coloured. The parameter defaults to false, so NRF Spectrum, Jam Detect
   and Channel Analyzer are unchanged and keep the theme ramp.

2. RF Waterfall is rebuilt on that component. It previously swept the CC1101 one
   pixel-column at a time, mapped RSSI through a hand-rolled rainbow with an
   inverted ramp (strong signals came out dark, the noise floor bright red),
   scrolled a manual framebuffer and drew its labels in literal TFT_* colours
   with no border, title or status bar. Now:

   • the top is SpectrumPlot's filled trace with a peak-hold line and animated
     noise floor; the drawn envelope eases toward each measurement (as Jam
     Detect does) so the trace glides instead of snapping
   • the waterfall below is fed by the same envelope, so the falls track the
     waveform above them, drawn with the new SDR colourmap
   • a five-tick frequency ruler and a status line naming the peak dBm and
     frequency, all theme-consistent
   • UP/DOWN (or PREV/NEXT) pan the whole window; the ruler and peak-hold reset
     to follow

   The sweep itself is fixed too. In the scan path setMHZ only rewrites the
   frequency, and the CC1101 needs a few milliseconds for the PLL/RSSI to settle
   before getRssi() is valid (the working RF RSSI view waits 5ms). The old
   per-pixel 100us wait was far too short, so a real signal never rose above the
   noise floor - the rainbow only ever showed noise. It now samples WF_BINS
   points with a real settle and interpolates the envelope across the plot, so a
   transmitter actually paints the trace and the waterfall.

Types of Changes

Bugfix (RSSI never settled, so nothing was detected; the colour ramp was
inverted) plus a UI rework onto the shared component and an opt-in rendering
option on that component. No behaviour change for existing SpectrumPlot callers,
no breaking change.

Verification

pio run -e m5stack-cardputer

Flash and open RF -> Waterfall on a device with a CC1101. Expect:

• a filled trace with peak-hold across the top, gliding as it updates
• a waterfall below in the SDR gradient (dark for the noise floor, climbing
  through blue/green/yellow to red/white on a strong signal) that follows the
  trace
• a frequency ruler and a status line naming the peak dBm and frequency
• UP/DOWN pans the band

Transmitting on a frequency inside the band should raise the trace at the
matching column and leave a bright streak falling down the waterfall.

NRF Spectrum, Jam Detect and Channel Analyzer should be visually unchanged
(they keep the theme ramp).

Testing

No unit test harness exists for this layer. Compiled clean for m5stack-cardputer.
Not yet verified on hardware by me - a reviewer with a CC1101 should confirm a
known transmitter paints the trace and that the pan controls track.

Linked Issues

None.

User-Facing Change

RF Waterfall now renders on the shared spectrum plot with a gliding peak-hold trace and an SDR-style colour waterfall, follows the theme for its chrome, and actually detects signals (the sweep now lets the CC1101 settle before reading RSSI).

Further Comments

WF_BINS (64) and WF_SETTLE_MS (3) trade sweep resolution against refresh rate;
they are #defines at the top of the run function so they are easy to tune on
hardware. The SDR gradient stops follow the common SDR waterfall colourmaps
(GQRX, SDR#, SDRPlay); index 0 is the panel background so a dead band blends
with the frame rather than showing a hard black bar.