test(rtabmap_slam): cover scan transform recovery in native tests

Replace the standalone reproducer with registered C++ GoogleTests for 2D and 3D scan recovery. Use real ROS inputs, separate DDS domains and bounded timeouts, and include rtabmap_slam in supported CI and coverage jobs.

Validated on Jazzy with RTAB-Map core 9ed83a7: both negative controls fail map recovery without the unlocks; both pass with the fix, followed by five repeated runs per case (10/10 passed). Broader distro CI has not been run locally.

.github/workflows/coverage.yml

@@ -60,7 +60,7 @@ jobs:
       # tested package depends on them (--packages-up-to), just not measured.
       - uses: ros-tooling/[email protected]
         with:
-          package-name: rtabmap_conversions rtabmap_util rtabmap_sync rtabmap_odom rtabmap_python
+          package-name: rtabmap_conversions rtabmap_util rtabmap_sync rtabmap_odom rtabmap_slam rtabmap_python
           target-ros2-distro: humble
           # RTAB-Map is installed in the image, not as an apt package, so rosdep
           # cannot resolve the key and must not try.
@@ -120,7 +120,7 @@ jobs:
           . /opt/ros/humble/setup.sh
           # C++ packages only -- rtabmap_python emits no .gcno for lcov to read,
           # and is measured by the coveragepy step below instead.
-          PKGS="rtabmap_conversions rtabmap_util rtabmap_sync rtabmap_odom"
+          PKGS="rtabmap_conversions rtabmap_util rtabmap_sync rtabmap_odom rtabmap_slam"
           # Baseline from the .gcno files. Without it a source file that no test
           # ever loaded is missing from the report altogether rather than
           # counted as 0%, which quietly inflates the result.