Exercise 2: Two Independent Bugs, One Integration Test — Possible Solution
====================================================================

THE SECOND BUG
------------------------------
    def apply_discount_dollars_BUGGY(price_dollars, discount_pct):
        return price_dollars - discount_pct  # subtracts the raw percentage number
                                              # instead of applying it as a percentage

RESULTS
------------------------------
    Expected (no bugs): $17.99

    One bug (buggy price fn only, correct discount fn):
      $1.80, PASSES: False

    Two bugs (buggy price fn AND buggy discount fn together):
      $-8.00, PASSES: False

Both scenarios fail the same single integration test. The test's own
pass/fail result and even its printed value give no direct signal
about how many distinct faults contributed - "$1.80, wrong" and
"$-8.00, wrong" are both just "the total is wrong," with no built-in
way to tell from the integration test alone whether one component or
several are at fault.

WHY THIS WORKS AS AN ANSWER
------------------------------
This confirms the chapter's own warn-box directly: an integration test
trades fault-localization precision for broader real-code coverage.
Here that tradeoff is made concrete in a new way - it's not just that
the test can't say WHICH stage is broken, it can't even say HOW MANY
stages are broken. Diagnosing that would require exactly what the
chapter's own pyramid argument recommends: falling back to the
individual unit tests for calculate_price_in_cents and
apply_discount_dollars, each of which would independently and
precisely report its own specific fault.