Add status attribute to output files for improved error handling and checkpointing

[Aaryan-549]

Description

This PR implements an MVP for checkpointing and error file tracking by adding a status attribute to output NetCDF files. This ensures that users and automated systems can easily determine whether a simulation completed successfully, is a checkpoint, or encountered an error condition.

Addresses: #1679

Problem

Currently, TORAX simulations that encounter errors (NaN detection, negative profiles, quasineutrality violations, or reaching minimum timestep) may not write output files consistently. This makes it difficult to:

• Debug long-running simulations that fail.
• Distinguish programmatically between completed and failed runs.
• Implement checkpointing for verification, surrogate training, or long-pulse scenarios (e.g., DEMO, STEP, ARC).

Solution

This implementation follows the design feedback from the issue, prioritizing simplicity and maximal reuse of existing infrastructure.

1. New SimStatus Enum (_src/state.py):

   • COMPLETED: Simulation reached t_final successfully.
   • CHECKPOINT: Intermediate checkpoint (reserved for V2/future use).
   • ERROR: Simulation stopped due to an error condition.

2. Status Attribute in Output (_src/output_tools/output.py):

   • All NetCDF output files now include a global status attribute.
   • This is automatically set based on the sim_error state.
   • Accessible via data_tree.attrs['status'].

3. Consistent Error File Writing (_src/simulation_app.py):

   • Output files are now explicitly written for ALL error conditions:
     • NAN_DETECTED
     • NEGATIVE_CORE_PROFILES
     • QUASINEUTRALITY_BROKEN
     • REACHED_MIN_DT
   • Files contain all valid timesteps captured up to the error.

Design Principles Followed

• Simplicity: Minimal changes with no complex logic introduced.
• Reuse: Uses the standard NetCDF file format and existing output_tools.
• No Breaking Changes: Purely additive; does not modify existing restart behavior or API.
• Single Format: Checkpoints/Error files use the exact same schema as regular output files.

Changes

• torax/_src/state.py: Added SimStatus StrEnum.
• torax/_src/output_tools/output.py: Added logic to inject the status attribute into the DataTree.
• torax/_src/simulation_app.py: Improved error logging and ensured file writing triggers on error states.
• torax/_src/output_tools/tests/output_test.py: Added comprehensive tests for the status attribute.

Testing

• Compilation: All files compile successfully.
• New Tests: Verified status attribute is correctly set to COMPLETED for successful runs and ERROR for failed runs.
• Regression: Existing tests pass; fully backward compatible.

Future Work (V2)

This MVP establishes the foundation for advanced checkpointing features planned for the future, including:

• Periodic checkpoint writing during simulation execution.
• Unification of Checkpoint and Restart APIs.
• Configurable checkpoint intervals and automatic cleanup.