Publish the afmotion image and pull from docker hub, and add failure reporting

[avhekkink]

Published the afmotion image as cemrg/afmotion:1.0 on Docker Hub and pointed the code at it. Also added an output to check for every step so that a step no longer falsely reports success when nothing has been written.

Additions

• RunAndCheckDockerAtrialStrainMotionSubcommand runs one afmotion subcommand and checks it.
  It replaces the per-step ExecuteTouch and DockerAtrialStrainMotion pairs, and names the step
  and the missing file in a warning dialogue.

• SetDockerImageAfmotion names cemrg/afmotion:1.0. afmotion was a bare local image name,
  so the pipeline ran on one machine. The image is now on Docker Hub, public — the September 2024
  build published without a change, digest sha256:6574a5bc…ade70c, 22 layers. A rebuild needs a
  VTK 9.3 source tree, and the tag the author used isn't recorded anywhere. Will look into this as a separate piece of work.

Edits

• DockerAtrialStrainMotion returns a verdict. It takes expectedOutput instead of returning
  void, runs the container as the host user, and reads the image name from GetDockerImage.

• GetDockerArguments now passes --workdir=/data, and this repairs UAC Stage 1. It passed no
  working directory before, so each container used the one it declares. cemrg/uac declares
  WorkingDir=/code, which root owns. Once PR Run the Docker containers as the host user #94 added --user, the host user could no longer
  write there, and read_array_fiber writes a temporary file by a relative path — so Stage 1 raised
  PermissionError and died before it wrote LSbc1.vtx. The container exited 0 regardless, so step
  10 reported four missing outputs and no reason. --workdir=/data points the working directory at
  the mount, which is what GetOpenCarpDockerCoreArguments already does with --workdir=/shared.

• Nine ExecuteTouch calls are gone, across AutoLandMark, UAC_Stage2, CreateModel,
  Registration, GenerateCellAreaStrains and JacobianThreshold. A pre-created empty file made
  the check pass on a step that wrote nothing; --user removes the reason those calls existed.

• CreateModel and Registration stop at the first failure. Each subcommand reads what the one
  before it wrote. Two checks can still pass on a partial run — generateCellAreaStrains and
  calcFiberStrains loop in bash with no set -e — and a comment marks that limit.

Impacted areas

Area	Files
Atrial Strain Motion plugin	AtrialStrainMotionView.{h,cpp}
CemrgCommandLine (shared module)	CemrgCommandLine.{h,cpp}

GetDockerArguments has four callers. Three run cemrg/uac and all gain the working-directory fix;
DockerCctaMultilabelSegmentation runs cemrg/ccta, which already declares WorkingDir=/data, so
nothing changes there. The Atrial Fibres plugin calls two of the three, so it had the same UAC
Stage 1 failure and gets the same fix. It is not re-run on this branch.

Testing

Verified on V-0004 and on the 2024 V-0005 regression fixture:

• Steps 1 to 9 pass on V-0004. UAC Stage 1 now writes its five .vtx files, owned by the host
  user — it failed before the working-directory fix.
• Step 9 runs against the published image through the panel, with afmotion:latest removed first so
  only cemrg/afmotion:1.0 could serve it.
• autoLM reproduces the 2024 reference after a fresh docker pull, md5
  14a5ed485da22b3e6a20009a26db8c54.
• MitkCemrgAppModule compiles and links.

Not verified:

• Steps 10 and 11 fail, and this branch does not cause it. openCARP rejects ellip_use_pt and
  parab_use_pt. docker.opencarp.org/opencarp/opencarp:latest is a development build, and
  openCARP removed both parameters after release v19.0. The same command fails as root.
• Steps 12 to 20, because step 11 blocks them.
• The Atrial Fibres plugin, after the shared GetDockerArguments change.
• Windows. GetDockerUserArguments returns nothing there and ExecuteTouch was already a no-op, so
  the checks behave differently by platform.

[alonsoJASL]

Just help me with that one comment on the line in the plugin's cpp:

class CemrgCommandLinel 

Otherwise it's ready.

[alonsoJASL]

Go ahead and merge