feat(operate): calibration gets its own pane (#108)

[pskeshu]

Judgement call, flagged

Kesavan raised on the walkthrough whether one embryo's calibration should apply to all of them, or each should carry its own. That is a question about the experiment, not the interface — so the pane is built per-embryo, matching what calibrate_embryo already does, and the question is left open rather than answered by a layout.

---

Calibrate was a button in the SPIM head's status bar, and its result appeared in a different tab entirely.

Ryan, 25:30: "after setting up the bottom camera and SPIM head, I feel like I'm not quite sure what to do in gently at this point."

There was no next step on screen, because the next step was a button in the corner of the step before. His own description of the order is now the pane order:

Bottom cam → SPIM head → Calibration → Acquisition

The pane

Declares its method before running it — Kesavan, same call: "it should show what is the method it is going to use to calibrate." One method on this rig, so it is named rather than offered as a choice, and the text says the part that matters: it needs the laser, because the LED alone gives a DIC-like image with no nuclei to find. That is the confusion #106 turns on, written where the button is.

It carries the Calibrate button, the selected embryo, the fit readout, and the focus-sweep montage — which previously appeared under the SPIM head while the result lived elsewhere. Control, evidence and result in one place.

Leaving the pane closes the LED. calibrateSelected never did, which is half of #106.

The rail says which embryos still need it

RosterPanel gains showFit, reading the same field the server-side gate checks (a finite non-zero slope_um_per_deg). The shared rail sits beside every pane, so CAL / UNCAL is visible wherever you are — an operator sees what a run will refuse before reaching Start rather than discovering it there.

Muted, not red: an uncalibrated embryo is one you haven't got to yet, not an error.

Also

PANE_ORDER is one list — it used to be spelled out in showPane and again in showPaneInitial, so adding a pane meant remembering both. And three stray double publishRoster() calls left by the roster rename are collapsed.

Verified live

Subtab order · pane shows while others hide · Calibrate and the montage host moved · SPIM bar carries neither · fit readout reads 96.4 µm/deg · R² 0.91 for a calibrated embryo and not calibrated for one without · rail badges track both.