Dev container: clone path, QT_XCB_NO_MITSHM, vim

[fjp]

Code fixes found while reviewing the comments in #105.

• setup.sh: repositories cloned into src/src/. vcs import --skip-existing src < src/diffbot/diffbot_dev.repos passed src as the base folder, but the paths in diffbot_dev.repos already start with src/. So rplidar_ros and remo_description ended up in ~/catkin_ws/src/src/. The build still worked because catkin searches subfolders. vcs import now runs without a base folder, as in the manual setup in the README.
• devcontainer.json: QT_XCB_NO_MITSHM=1 replaces QT_X11_NO_MITSHM=1. The old variable came from the ROS Docker GUI tutorial, but only Qt 4 reads it; Noetic's Qt 5.12 reads QT_XCB_NO_MITSHM. Without it, Qt 5 tries to share memory with the X server (MIT-SHM). That fails because the container has its own IPC namespace; Qt then gives up and draws normally.
• Dockerfile: vim next to nano. The ARG lines for the user's name and IDs sit next to the step that creates the user. find no longer filters .git, which .dockerignore already keeps out of the build context.
• host-x11.sh: the script's folder gets its own variable, script_dir.

A container created before this change has the two repositories in src/src/. Rebuild it (Dev Containers: Rebuild Container) before running the new vcs import command in it, or catkin finds each package twice.

The docs get the same vcs import fix and a paragraph on QT_XCB_NO_MITSHM in a separate PR in ros-mobile-robots.github.io.

Testing

• The image builds. vim is installed, the user ros has UID 1000, and the manifests stage copies the same 10 package.xml files as before.
• setup.sh in a container from the new image, with the clone mounted as in devcontainer.json: the repositories land in src/rplidar_ros and src/remo_description, there's no src/src, and all 13 packages build. A second run skips the existing repositories, builds again, and leaves a single source line in .bashrc.
• MIT-SHM: an Xvfb X server in one container and a Qt 5 window in another, sharing only the X socket folder, like the host and the dev container. With QT_X11_NO_MITSHM=1, Qt's debug log is the same as with no variable: xcb_shm_attach() failed, then disabling SHM. With QT_XCB_NO_MITSHM=1, Qt doesn't try. In all three cases the X server shows the window correctly.
• host-x11.sh called from another folder still writes .x11/xauth next to itself, with modes 700 and 600, and turns a test cookie into a ffff (wild) entry. Without an Xauthority file, as on WSLg, it leaves the file empty and exits with 0.