Evaluate Smartspim notebook registration accuracy with landmark metric

## Background

[LightsheetRegistration.ipynb](https://github.com/InsightSoftwareConsortium/itk-dreg/blob/main/examples/LightsheetRegistration.ipynb) provides an example workflow for multistage registration of SmartSPIM data to the Allen Institute Common Coordinate Framework (CCF) atlas.

## Current Behavior

The lightsheet example notebook evaluates registration results via the voxel difference image between the resampled moving image and the fixed image. Since registration is cross-modal, we expect large voxel differences, so it is difficult to know how far the registration result is from a "perfect" registration.

## Request Behavior

The Allen Institute has annotated the 652506 SmartSPIM sample and the CCF atlas with point landmarks. We should evaluate registration results as follows:
1. Transform SmartSPIM landmarks by the output forward transform from registration
2. Compute distance statistics between the transformed SmartSPIM landmarks and reference CCF atlas.