fix(divan_compat): do not capture drop with perf

[not-matthias]

No description provided.

[Copilot]

Pull Request Overview

This PR fixes performance measurement accuracy by moving the BenchGuard to only capture the actual benchmark execution time, excluding the overhead from input/output dropping operations.

• Removed the global BenchGuard that was capturing the entire benchmarking loop including drop operations
• Added localized BenchGuard instances around the core timing sections to measure only the benchmark execution
• Added a test benchmark to demonstrate the issue with large input drops

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
crates/divan_compat/divan_fork/src/bench/mod.rs	Moved BenchGuard placement from outer loop to inner timing sections to exclude drop overhead
crates/divan_compat/benches/drop_example.rs	Added test benchmark with large input that has expensive drop to verify the fix
crates/divan_compat/Cargo.toml	Added bench configuration for the new drop_example benchmark

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[codspeed-hq]

CodSpeed WallTime Performance Report

Merging #118 will degrade performances by 16.49%

_{Comparing cod-1344-investigate-benchmark-time-in-drop_in_place-and-perf (b02554d) with main (0565866)}

Summary

⚡ 14 improvements
❌ 21 regressions
✅ 119 untouched
🆕 11 new

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

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
❌	large_drop	163.4 µs	168.9 µs	-3.27%
❌	large_setup	14 ns	15 ns	-6.67%
❌	from_elem[16384]	867 ns	948 ns	-8.54%
❌	from_elem[2048]	243 ns	258 ns	-5.81%
❌	from_elem[8192]	514 ns	534 ns	-3.75%
❌	from_elem_decimal[2048]	248 ns	264 ns	-6.06%
⚡	bench_array1[10]	10 ns	9 ns	+11.11%
⚡	bench_array2[10]	10 ns	9 ns	+11.11%
❌	init_array[1000]	1.8 µs	2 µs	-9.76%
⚡	fibo_10	2 ns	1 ns	×2
🆕	bench_large_input	N/A	50.5 ms	N/A
🆕	bench_large_input_local	N/A	50.4 ms	N/A
🆕	fib_in_thread[30]	N/A	6.8 ms	N/A
🆕	fib_in_thread[31]	N/A	10.9 ms	N/A
🆕	fib_in_thread[32]	N/A	17.5 ms	N/A
🆕	fib_in_thread_bench[30]	N/A	6.8 ms	N/A
🆕	fib_in_thread_bench[31]	N/A	10.9 ms	N/A
🆕	fib_in_thread_bench[32]	N/A	17.5 ms	N/A
🆕	fib_in_thread_bench_local[30]	N/A	6.8 ms	N/A
🆕	fib_in_thread_bench_local[31]	N/A	10.9 ms	N/A
...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

[codspeed-hq]

CodSpeed Instrumentation Performance Report

Merging #118 will degrade performances by 6.78%

_{Comparing cod-1344-investigate-benchmark-time-in-drop_in_place-and-perf (b02554d) with main (0565866)}

Summary

⚡ 5 improvements
❌ 5 regressions
✅ 157 untouched
🆕 16 new

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

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
❌	sleep_100ms	595.3 ns	624.4 ns	-4.67%
❌	sleep_10ms	566.1 ns	595.3 ns	-4.9%
❌	sleep_1ms	566.1 ns	595.3 ns	-4.9%
❌	sleep_50ms	595.3 ns	624.4 ns	-4.67%
⚡	bench_array1[10]	254.2 ns	225 ns	+12.96%
⚡	bench_array2[10]	254.2 ns	225 ns	+12.96%
⚡	mut_borrow	924.2 ns	895 ns	+3.26%
🆕	bench_large_input	N/A	1.6 ms	N/A
🆕	bench_large_input_local	N/A	1.5 ms	N/A
🆕	sleep_100ms	N/A	566.1 ns	N/A
🆕	sleep_100ms_with_custom_sample	N/A	566.1 ns	N/A
🆕	sleep_10ms	N/A	566.1 ns	N/A
🆕	sleep_1ms	N/A	566.1 ns	N/A
🆕	sleep_50ms	N/A	566.1 ns	N/A
🆕	fib_in_thread[30]	N/A	9.7 ms	N/A
🆕	fib_in_thread[31]	N/A	15.8 ms	N/A
🆕	fib_in_thread[32]	N/A	25.5 ms	N/A
🆕	fib_in_thread_bench[30]	N/A	9.7 ms	N/A
🆕	fib_in_thread_bench[31]	N/A	15.8 ms	N/A
🆕	fib_in_thread_bench[32]	N/A	25.5 ms	N/A
...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

[not-matthias]

Closing in favor of #121