Preserve course subjects across discovery syncs and document the precedence rule

[Copilot]

Course subjects were being populated through two discovery sync paths with different overwrite behavior, which led to inconsistent categories/subcategories across otherwise similar courses. This change aligns subject handling across incremental and full metadata updates, and captures the rule in an ADR.

• Keep subjects fresh during incremental syncs

  • Add subjects to COURSE_FIELDS_TO_PLUCK_FROM_SEARCH_ALL.
  • Existing course records now pick up subject changes from /api/v1/search/all/ without replacing the rest of the stored full-course payload.

• Prevent empty full-course payloads from erasing known-good subjects

  • Update _update_single_full_course_record() so an empty or missing subjects value from /api/v1/courses/ does not overwrite an existing non-empty value.
  • Non-empty subjects from the full payload still replace the stored value.

• Document the decision

  • Add docs/decisions/0010-preserving-course-subjects-across-syncs.rst.
  • The ADR describes:
    • the inconsistent behavior
    • the split between /search/all/ and /courses/
    • the field-level precedence rule for subjects
    • the tradeoff of preserving known-good values over clearing them on empty full payloads

existing_subjects = metadata_record._json_metadata.get('subjects')
metadata_record._json_metadata.update(course_metadata_dict)

if existing_subjects and not metadata_record._json_metadata.get('subjects'):
    metadata_record._json_metadata['subjects'] = existing_subjects