Server: use forkserver as worker context

Server: detach predicted FoamData objects before postprocessing

Server: convert porous ids to numpy array when postprocessing

server: remove models lock

server: run data generation and postprocessing in own processes

server: add process pool to app

server: initialize app models outside lifespan function

server: add postprocess function

preprocessing: pass input data to create_session_folders

server: override meta files in generate_f

server: always use Generator2DVariable instead of Generator2DFixed

server: load PI-GANO models

Server: handle case generation process failure

Server: use rich for prediction progress

server: update imports to new submodule project structure

App initial commit