Trigger test result values off by 1

[JasonYeMSFT]

I ran the trigger tests of the code-explainer example. There are 8 test cases that should trigger the skill and 8 that would otherwise. None of the executions triggered the skill. I expected the result to report 8 FN and 8 TN but it reported 7 FN and 7 TN, both off by 1.

[spboyer]

Fixed in PR #193 (merged). The root cause was ComputeTriggerMetrics using confidence-weighted sums for integer TP/FP/TN/FN counts — medium-confidence prompts (0.5 weight) made counts fractional. Integer counts now always use whole numbers; weighted values are preserved for precision/recall/F1.

@JasonYeMSFT — resolved on main.

[spboyer]

Resolved in v0.25.0 via PR #193. The confidence-weighted sums were incorrectly used for integer TP/FP/TN/FN counts — now uses actual result counts for integers while preserving weighted values for precision/recall/F1.

@JasonYeMSFT