fix: skip warmup for embedded benchmarks

[not-matthias]

Replace the inline isWarmupFork() check inside runBenchmark's
measurement loop with an explicit isWarmup flag computed in doSingle
from !mode.doMeasurement() || isWarmupFork().

Groundwork for COD-2548 (warmup leaking into flamegraphs for <100ns
benchmarks): centralises the warmup-vs-measurement decision so the
outer startBenchmark/stopBenchmark scope can be tightened next.

[codspeed-hq]

Merging this PR will degrade performance by 12.82%

⚡ 4 improved benchmarks
❌ 2 regressed benchmarks
✅ 119 untouched benchmarks
⏩ 9 skipped benchmarks¹

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	findHighestSetBit[255]	47 ns	29 ns	+62.07%
⚡	fibonacciBottomUp[20]	1.4 µs	1.2 µs	+17.63%
⚡	fibonacciBottomUp[10]	500 ns	441 ns	+13.38%
⚡	dualPivotQuickSort[100]	14.4 µs	11.1 µs	+29.49%
❌	timSort[100]	9.5 µs	10.9 µs	-12.82%
❌	timSort[1000]	149.5 µs	168.6 µs	-11.34%

---

_{Comparing cod-2548-investigate-warmup-included-in-flamegraphs-in-100ns (c6a5576) with main (ff83f3b)}

Footnotes

1. 9 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩