fix(hook): report Windows pointer motion by differencing the cursor point

[Stanley5249]

Summary

WH_MOUSE_LL never emitted MouseEvent::Moved, so SwipeAccumulator saw no
travel during a gesture hold and a swipe could never commit on Windows — only the
plain click on release fired. This differences the cursor position each
WM_MOUSEMOVE against the previous one, giving the accumulator the relative delta
it expects, the same shape as macOS MOUSE_EVENT_DELTA_X and Linux REL_X.

Changes

openlogi-hook

• windows.rs: motion_delta() differences MSLLHOOKSTRUCT.pt against a
  Cell<Option<POINT>> thread-local; WM_MOUSEMOVE translates to
  MouseEvent::Moved. No synchronization — WH_MOUSE_LL callbacks run on the
  installing thread.
• Injected moves stay filtered from the event stream but do advance the baseline,
  so a cursor jump made by another process can't be reported as the user's travel.
• Tests: successive differencing (no-baseline first move, zero-delta repeat) and
  the injected-baseline case.

Known limitation

The OS clamps MSLLHOOKSTRUCT.pt to the desktop bounds, so a swipe running into a
screen edge stops producing deltas and may not reach the threshold. Documented on
motion_delta. #472 is the alternative that avoids it, at the cost of ~350 lines
of raw-input FFI; these two are alternatives, not a sequence.

Testing

• cargo clippy -p openlogi-hook --all-targets — clean
• cargo test -p openlogi-hook — 11 passed
• Not runtime-tested on hardware.

Note: the Windows CI job fails until #458 lands — openlogi-hid does not build
without the Win32_System_IO feature.

Fixes #471

[greptile-apps]

Greptile Summary

Adds Windows pointer-motion translation for gesture swipes.

• Differences consecutive WH_MOUSE_LL cursor positions into MouseEvent::Moved deltas.
• Advances the cursor baseline for filtered injected movement.
• Documents edge-clamped motion and adds unit tests for differencing and injected movement.

Confidence Score: 3/5

The first-movement handling should be fixed before merging because it can still turn a valid Windows swipe into a click.

A gesture can begin while the thread-local cursor baseline is empty, causing the first and potentially only threshold-crossing movement to be discarded before it reaches the swipe accumulator.

Files Needing Attention: crates/openlogi-hook/src/windows.rs

Important Files Changed

Filename	Overview
crates/openlogi-hook/src/windows.rs	Adds Windows motion differencing and injected-event baseline handling, but drops the first movement even when it occurs during an active gesture.
crates/openlogi-hook/src/lib.rs	Documents that Windows pointer motion is edge-clamped.

Sequence Diagram

sequenceDiagram
  participant OS as Windows WH_MOUSE_LL
  participant Hook as translate_event
  participant Delta as motion_delta
  participant Swipe as SwipeAccumulator
  OS->>Hook: WM_MOUSEMOVE with cursor point
  Hook->>Delta: Difference from LAST_POINT
  alt Baseline exists and position changed
    Delta-->>Hook: delta_x, delta_y
    Hook->>Swipe: MouseEvent::Moved
  else First observed point
    Delta-->>Hook: None
    Note over Swipe: Movement is not accumulated
  end

Loading

_{Reviews (1): Last reviewed commit: "fix(hook): report Windows pointer motion..." | Re-trigger Greptile}

[greptile-apps]

First gesture movement is dropped

When a gesture button is pressed before the hook has observed any pointer movement, motion_delta uses the following WM_MOUSEMOVE only to initialize LAST_POINT, causing that movement to be omitted from the swipe accumulator and a valid swipe to resolve as the fallback click.

Knowledge Base Used: openlogi-hook: OS-level input hooking