Compute the Gabor spectrogram in one place #18

What is wrong

croak computes a Gabor spectrogram in two places with different choices:

• croak.processing.spectrogram(result, ...) — the retrieval and dispersion views: the gate centres span half the time window, the gate is a twentieth of the window wide (FWHM), on a RetrievalResult only;
• the synthetic generator's preview, now croak.plotting.screen._gabor — gate centres over the pulse with a 30 % margin, a Gaussian σ of a thirtieth of that span (at least four samples), on any field.

Why it matters

Two implementations of one quantity drift apart, and the generator's version is not reachable from scripts. A single function that takes a field (grid, spectrum, ω₀) with the gate and window as explicit, documented choices would serve both, with processing.spectrogram(result) as a thin wrapper.

What done looks like

One spectrogram function on a field, tested against an analytic case (a linearly chirped Gaussian, whose spectrogram ridge has a known slope), used by both callers; the generator's preview unchanged in what it shows (zoomed to the pulse).

Context

Noticed while moving the generator's preview to the screen-data layer in #16, which kept its computation as it was to stay in scope. Read, not run.

Suggested labels: refactor, area: gui, priority: low.