[App quality] Make zoom-to-physical-size correct per display and calibratable

[mberrys]

Part of #155.

Problem

The product is named for a magnifier and the vision doc commits it to being "exact about geometry". The zoom-to-physical-size mapping is not exact, and on the multi-monitor setups that print shops actually use it is wrong.

Pdf4QtLibWidgets/sources/pdfdrawspacecontroller.cpp:554-570:

QScreen* primaryScreen = QGuiApplication::primaryScreen();
QSizeF physicalSize = primaryScreen->physicalSize();
QSizeF pixelSize    = primaryScreen->size();
m_pixelPerMM = pixelSize.width() / physicalSize.width();

Three defects follow:

1. Always the primary screen. The scale factor is derived from primaryScreen() regardless of which display the window is on. A laptop panel plus an external 4K monitor — the standard studio setup — gives a wrong millimetre-per-pixel ratio on the secondary display, and nothing recalculates when the window is dragged between screens. "100%" is then not 100% and a hairline measured on screen is not the hairline on press.
2. physicalSize() is EDID-reported and frequently wrong. Monitors misreport, projectors and VMs report nothing useful, and remote sessions synthesise values. There is no operator calibration path (Acrobat exposes one; a print tool needs one).
3. Only the primary screen's width is used — physicalSize.height() and non-square pixel aspect are ignored.

Separately, Pdf4QtLibGui/main.cpp:33 sets Qt::AA_DisableHighDpiScaling, which no other entry point does. Whatever that surface is for, it renders at 1× on a high-DPI display while the shipped Editor does not — an inconsistency that should be resolved rather than left in the tree.

Scope

1. Derive the pixel-per-millimetre ratio from the screen the window is currently on (QWidget::screen()), and recalculate on QWindow::screenChanged and on QScreen::physicalDotsPerInchChanged.
2. Add an operator display calibration: a settings surface where the user matches an on-screen ruler to a physical one, persisted per display (keyed by screen name/serial), overriding the EDID-derived value when set.
3. Show the effective source of the current scale — reported by display, or calibrated by operator — wherever "actual size" is claimed, so an operator can tell whether to trust it.
4. Handle non-square pixel aspect, or state explicitly that it is unsupported and detect it.
5. Resolve AA_DisableHighDpiScaling in Pdf4QtLibGui/main.cpp: remove it, or document why that surface differs.
6. Confirm the QPainter renderer path (pdfdrawwidget.cpp:848-853) reaches device-pixel resolution on fractional scale factors. The Blend2D path handles device pixel ratio explicitly (pdfdrawwidget.cpp:819-830); the QPainter path does not, and the two should be verified to produce the same measured output.

Acceptance criteria

• Dragging a window between displays with different DPI updates the scale factor, and a 100 mm object measures 100 mm on both, within the calibration tolerance.
• With calibration set, "actual size" matches a physical ruler within 0.5 mm over 100 mm on the calibrated display.
• Calibration persists per display and survives disconnect/reconnect of an external monitor.
• The Blend2D and QPainter engines produce the same measured page geometry at the same zoom on the same display.
• No entry point disables high-DPI scaling without a recorded reason.

Test strategy

• Unit: scale-factor computation against synthetic QScreen geometries, including a two-screen configuration where primary and current differ, and a screen reporting zero physical size.
• Integration: render a known-size fixture at 100% and assert the pixel extent equals the expected physical extent for a given DPI.
• Manual, on the named reference device: physical-ruler check on primary and secondary displays, before and after calibration.

Dependencies

• Coordinates with [0.2.0] Add spatial-index hit testing and precise drag/zoom semantics #145 (precise drag/zoom semantics) — that issue owns the interaction; this one owns the unit mapping underneath it.
• Any measurement surface built on the Dimensions plugin inherits this scale factor and should be re-verified after the fix.

[mberrys]

Consolidated under #536 (1.x expansion umbrella) per the 2026-09-06 release-train consolidation. Reopen and promote with a milestone when this becomes active scope.