feat(api): adjust AI Risk score display

[hoangsonww]

This PR adjusts AI risk score display to correctly preserve the randomization logic (so ~50% of case actually has the risk score).

[Copilot]

Pull Request Overview

This PR refines how the AI CRC risk score is surfaced by prioritizing a literal CSV-provided value, falling back to database observations, and removing default “N/A” when no reference exists.

• Tracks whether CSV provided a literal CRC score or just toggled the section
• Fetches and renames observations when only toggled
• Omits the CRC score node entirely if CSV had no relevant entries

Comments suppressed due to low confidence (5)

src/cases/service/case_service.py:297

• [nitpick] The variable name csv_crc_score_leaf could be clearer about its role as the literal score text. Consider renaming it to something like csv_literal_crc_score_text for better readability.

csv_crc_score_leaf: str | None = None

src/cases/service/case_service.py:298

• [nitpick] Consider renaming old_crc_toggle to a more descriptive name such as crc_section_toggle or csv_crc_section_toggle to clarify that it represents whether the CSV toggled the CRC risk section.

old_crc_toggle = False

src/cases/service/case_service.py:370

• Add unit tests covering the new branches: when a literal CSV score is provided (csv_crc_score_leaf), when only the toggle is set (old_crc_toggle), and when neither flag is present to confirm no CRC node is appended.

if csv_crc_score_leaf:

src/cases/service/case_service.py:377

• Extract the magic number 45614722 into a named constant (e.g., CRC_CONCEPT_ID) to avoid hard-coded values and improve maintainability.

[45614722]

src/cases/service/case_service.py:409

• [nitpick] The docstring lists six steps, but this inline comment labels step 7. Consider aligning the numbering between the docstring and inline comments to avoid confusion.

# --- 7) Return the assembled Case object ---