fix(peaks): zoom-scaled apex snap and Shift free placement

[Limdongcheng]

Root cause

Manual picking snapped to the tallest local maximum inside a window fixed at 1.5% of the full x span, regardless of zoom — so next to a strong line a weak line could never be picked: zooming in sharpened the click but not the window.

Fix

• The apex search window is now derived from a fixed screen radius (12 px) at the current zoom, so zooming in narrows it in step with the click precision the zoom affords.
• Tallest-in-window is kept (over nearest-local-maximum) so zoomed-out clicks land on real peaks instead of the nearest noise wiggle.
• Shift+click skips the snap and places the mark on the nearest sample — the escape hatch for shoulders the apex search refuses to land on. The hover preview resolves through the same path, modifier included.

API: Trace1d::snap becomes snap_within(x, half_width) + nearest_sample, unified under pick(x, ManualPeakSnap); add_manual_peak takes the snap explicitly.

Range drag-picking (pick_in_range) keeps its 3-sigma prominence floor: it is scoped to the dragged window with MAD-based noise, which peaks inflate very little, and the reported failure was click-snap, not range picking.

Docs & tests

New peaks_tests.rs pins the narrow-window weak-apex pick, the wide-window tallest pick, free placement, and the no-apex fallback. The peak-picking guide (EN + zh-CN) documents the zoom-scaled snap and Shift placement. cargo pr-check passes.

---

Part 4 of the stack, based on feat/plot-fit-shortcuts (#4).

[github-actions]

Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. You can sign the CLA by just posting a Pull Request Comment same as the below format.

---

I have read the CLA Document and I hereby sign the CLA

---

_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}