Support padding in RLSSM #870
Description
If padding support is needed, we could approach it two ways:
Option 1: User-side preprocessing
Provide a separate utility tool for users to pad/mask their data before passing it to the model. This keeps the core codebase simple while giving users control over how padding is handled for their specific research context.
Option 2: Built-in support (separate PR)
Add padding as a core feature, which would require:
- General masking API across RL models (not model-specific)
- Updated validation to conditionally allow non-uniform trials when masks are provided
- Documentation on RL state propagation implications
- Comprehensive tests
Recommendation: Option 1 avoids adding complexity to the core codebase while still supporting use cases that need padding.
Originally posted by @cpaniaguam in #864 (comment)