Award no loyalty points on refunds

[sergical]

pointsFor now returns 0 for a negative total, so a refund no longer takes points away from the customer. There is one new test for a refund.

Created with Claude Code

[github-actions]

Verdict: Approve — small, correct, well-tested change that makes refunds (negative totals) award zero loyalty points.

Correctness

• Logic is sound: the early-return guard (totalCents < 0 → 0) correctly short-circuits before the rounding math, and doesn't affect existing behavior for totalCents >= 0.
• New test (pointsFor(-10_000) === 0) is meaningful and passes.
• All tests pass: 17 passed, 0 failed (2 suites run).
• No edge cases missed for the stated scope (e.g., -0 and non-integer cents still behave as before; boundary at 0 is unaffected).

Style (minor, non-blocking)

• The updated doc comment packs two unrelated facts (rounding rule + refund rule) into one run-on sentence. Consider splitting into two sentences/lines for scannability, e.g.:

  /**
   * One and a half points per whole dollar spent, rounded once after the multiplier.
   * Negative totals (refunds) always earn zero points.
   */
  

• The guard clause's intent (negative = refund) isn't obvious from the code itself — an inline comment near the if would help readers who skip the docblock.
• The new test only checks one refund value (-10_000); consider also covering a boundary case like -1 for consistency with the suite's existing boundary-focused style.
• The new test is appended at the end of the file; grouping it next to the other "awards nothing for …" tests would improve readability/organization.

None of the style points are blocking — the change is small, correct, and adequately tested.