feat: normalize benchmark mode to AverageTime under CodSpeed

[not-matthias]

Summary

• Normalizes all benchmark modes to AverageTime when CodSpeed instrumentation is active
• Handles three paths: CLI -bm flag overrides, Mode.All expansion, and annotation-level modes (e.g. @BenchmarkMode(Mode.Throughput))
• Prints an informational message when normalization occurs so users know their mode was overridden

Resolves COD-2446

[linear]

COD-2446 Normalize `BenchmarkMode` to `SampleTime`

JMH benchmarks can use different BenchmarkMode values such as average time, sample time, and throughput. The current implementation derives statistics from the selected mode results, which means results can’t be merged reliably across modes.

For CodSpeed’s Java walltime support, the benchmark mode should be normalized instead of only warning and continuing. A first step is to override unsupported modes to SampleTime, since it provides the most precise basis for the statistics currently reported.

This should preserve plug-and-play behavior while avoiding incompatible result shapes.

[Copilot]

Pull request overview

This PR updates the JMH runner to ensure that when CodSpeed instrumentation is active, benchmarks always execute using Mode.SampleTime, regardless of whether the mode was set via CLI (-bm), Mode.All, or per-benchmark annotations, and emits an informational message when an override occurs.

Changes:

• Normalize CLI-specified benchmark modes (-bm) to SampleTime under CodSpeed instrumentation.
• Normalize Mode.All expansion and any non-SampleTime annotation modes to SampleTime under CodSpeed instrumentation.
• Print informational messages when normalization occurs to make the override visible to users.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codspeed-hq]

Merging this PR will improve performance by 38.73%

⚡ 2 improved benchmarks
✅ 123 untouched benchmarks
⏩ 9 skipped benchmarks¹

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	insertionSort[100]	17.2 µs	12.4 µs	+38.73%
⚡	timSort[1000]	168.4 µs	147.2 µs	+14.45%

---

_{Comparing cod-2446-normalize-benchmarkmode-to-sampletime (8cb09f5) with main (4b7a767)}

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. ↩

[GuillaumeLagrange]

LGTM, although we need to clear up the Average/Mean discussion that we had in the daily this morning before merging this

[not-matthias]

@GuillaumeLagrange I added a new CodSpeed benchmark mode which now saves iters_per_round and times_per_round_ns rather than back-calculating it from the avg value.

[GuillaumeLagrange]

lgtm