Add defensive assertions for required keys in preprocessing

[hgbytes]

Add defensive assertions to preprocessing functions

Problem

compute_inputs_and_labels() and compute_token_types() access dictionary keys without validation, risking uncaught KeyError exceptions if called with malformed input. While currently safe (keys are always provided by callers), this creates a latent bug that surfaces only during future refactoring or code reuse.

Solution

Add assertions at function entry points to validate required keys:

• compute_inputs_and_labels(): Assert "for_observation" and "for_target" exist
• compute_token_types(): Assert "inputs" and "labels" exist

This follows the codebase's existing validation pattern (assertions throughout).

Changes

• Added 2 assertions to compute_inputs_and_labels() (lines 59-60)
• Added 2 assertions to compute_token_types() (lines 85-86)
• Updated docstring in compute_token_types() to document required input format
• Added preprocessing_test.py with 11 unit tests covering:
  • Valid inputs produce correct output
  • Missing keys raise clear AssertionError
  • Extra keys are preserved
  • Input dict is not mutated

Impact

• No breaking changes: All existing code passes (keys are always provided)
• Infrastructure only: No changes to model logic
• Improves maintainability: Catches misuse early with clear error messages
• Test coverage: 11 new tests verify assertions work correctly

Testing

• Updated unittest covers valid inputs, missing-key cases, and preservation of extra fields.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.