Add diurnal fingerprint plot to the readout panel (#28)

[aivanchenk]

Closes #28.

What

Adds a fingerprint plot: the flux-community view of a diurnal cycle. An hour-of-day (0-23, 0 at the bottom) by day heatmap of the pixel's flux, colored by a diverging scale centered on zero so midday uptake (negative NEE) separates cleanly from nighttime/seasonal release (positive NEE). Reference: Fig. 4 of the FLUXCOM-X data paper (hal-04138821 / Biogeosciences 21, 5079, 2024).

It sits behind a Line | Fingerprint toggle in the readout panel and shares the same series, history slider, and loading/error states as the existing line chart.

How

• Pure consumer of the existing time series. The reader already returns the full flat [day x 24] Float32Array per pixel (the hour axis is only collapsed downstream in TimeSeriesPlot via dailyMeanSeries). The fingerprint plot indexes values[day * 24 + hour] instead of averaging, so there is no data-layer change.
• Canvas cell grid, no new dependency. Draw cost is bounded by the destination pixel width (nearest-sampled along the day axis), so multi-year windows (~184k cells) render without jank.
• Diverging color scale keyed on isLight, matching the existing timeSeriesChartTheme pattern. NaN cells (ocean / missing pixels) render transparent so gaps read as gaps.

New / changed files

• src/lib/map/fingerprintScale.ts (new) + fingerprintScale.test.ts (new) - diverging scale, symmetricAbsMax (NaN-safe), axis tick helpers.
• src/components/map/FingerprintPlot.tsx (new) - canvas heatmap with hour/day axes and a color legend.
• src/components/map/FingerprintPlotLoading.tsx (new) - skeleton matching the canvas footprint.
• src/components/map/MapReadout.tsx - Line | Fingerprint toggle; renders the matching plot/skeleton per view.

Verification

• Unit tests for the scale + helpers (npx vitest run: 53 pass, 8 new).
• npx tsc --noEmit clean on touched files; npx next build smoke test passes.
• Manual: selected a land pixel, confirmed the midday uptake band is visible with hour 0 at the bottom, the diverging legend is symmetric around 0, and the ramp updates in both light and dark themes. Switching back to Line leaves the existing chart unchanged.

🤖 Generated with Claude Code

[github-actions]

🚀 Vercel preview deployed

Preview URL: https://earth-prints-gesuh3t9b-anastasiia-s-projects10.vercel.app

[lazarusA]

• Usually the axes are flipped, could we add an option to do that?
• Also, we will need another option to single-out years, like a button group, showing available years, and from there click select just one to see.
• Also, we need a hover tooltip to show values in the finger print.

[aivanchenk]

@lazarusA thanks for the review, all three points are addressed in 1000c20:

• Flip axes — a toggle that swaps the hour and day axes.
• Single out a year — a button group of the calendar years the loaded window spans (All 2016 … 2021); pick one to view just that year. The color scale stays global so years remain comparable. Axis labels now use real calendar dates, derived from the store's time origin (days since 2001-01-01).
• Hover tooltip — reads the date, hour, and value under the cursor (e.g. 2019-07-11 · 17:00 · -0.21 umolCO2 m-2 s-1).

All view logic is client-side over the already-loaded series, so no extra fetch. Also note the colors changed since your screenshot: the ramp is now the geoscience-standard vik diverging scale (blue = uptake, red = release), so the Vercel preview will look different from the earlier teal/orange.

[lazarusA]

there is not new preview, because build was skipped.

[aivanchenk]

Heads-up: CI is green again on the latest commit (e4a492a) and the Vercel preview has been rebuilt and redeployed, so it now reflects the vik ramp plus the three new controls (flip axes, year single-out, hover tooltip).

• Lint + build + test: pass
• Preview deployment: pass → https://earth-prints-gesuh3t9b-anastasiia-s-projects10.vercel.app/map

(The earlier preview had skipped because a lint error failed the build, and the branch had also drifted into a conflict with main after #67 merged; both are resolved now.)

[lazarusA]

I just noticed, that when displaying more than one year, tick labels should be just year, not month, because months are now actually within each year.