feature : Add proprietary support for directory-based 2D image stacks across all workflows

Description

pytc-client already has broad support for native 3D volume inputs. However, many biomedical imaging datasets are stored as directories of 2D image slices (e.g. PNG/TIFF stacks), which currently require manual conversion into 3D volumes before use.

This issue proposes adding native support for folder-based 2D image stacks, bringing them to feature parity with existing 3D volume workflows and reducing unnecessary preprocessing steps.

Implementation Notes

• Should just be a check of the proposed input directory (is it a directory?) -> an image.io command

Acceptance Criteria

• Folder-based 2D image stacks load correctly in visualization
• 2D stacks can be used for inference without manual conversion
• Training pipelines accept 2D stacks with no workflow-specific hacks
• Documentation added describing supported directory structures and assumptions

Start by locating the existing 3D volume input handling and the image.io command mentioned in the implementation notes. Trace how inputs reach visualization, inference, and training workflows, then verify that directory-based 2D stacks load in each path and document the supported directory structures and assumptions.