Add Akka.Streams throughput benchmarks

[Aaronontheweb]

Summary

• Add StreamThroughputBenchmarks with 4 benchmarks measuring element throughput for different stream topologies
• Add ThroughputBenchmarkConfig that includes Req/sec column for easier comparison

Benchmarks Added

Benchmark	Topology	Purpose
Linear_Pipeline_Throughput	Source → 2 Selects → Sink	Most common topology baseline
Deep_Pipeline_5_Stages_Throughput	Source → 5 Selects → Sink	Amplifies per-stage overhead
Merge_Fan_In_Throughput	2 Sources → Merge → Sink	Fan-in topology baseline
Batch_Fan_In_Throughput	Source → Batch → SelectMany → Sink	Batching topology baseline

Sample Results (VM, expect better on bare metal)

Method	Mean	Req/sec	Allocated
Linear_Pipeline_Throughput	98 ns	10.2M	73 B
Deep_Pipeline_5_Stages_Throughput	213 ns	4.7M	145 B
Merge_Fan_In_Throughput	82 ns	12.2M	49 B
Batch_Fan_In_Throughput	183 ns	5.5M	201 B

Test plan

• Benchmark project builds
• Benchmarks execute successfully (dry run verified)
• Run on bare metal for canonical baseline numbers

[Aaronontheweb]

Akka.Streams Throughput Benchmark Results — old-gpu Baseline

Machine: Intel Core i7-6700K CPU 4.00GHz (Skylake), 1 CPU, 8 logical / 4 physical cores
OS: Linux Ubuntu 24.04.2 LTS (Noble Numbat)
Runtime: .NET 10.0.5 (10.0.5, 10.0.526.15411), X64 RyuJIT x86-64-v3
SDK: 10.0.201
BenchmarkDotNet: v0.15.4
GC: Concurrent Workstation

Method	Mean	Error	StdDev	Req/sec	Gen0	Gen1	Allocated
Linear_Pipeline_Throughput	106.33 ns	2.030 ns	2.172 ns	9,404,712.54	0.0173	-	73 B
Deep_Pipeline_5_Stages_Throughput	227.15 ns	4.528 ns	7.050 ns	4,402,393.09	0.0344	-	145 B
Merge_Fan_In_Throughput	89.43 ns	1.731 ns	2.370 ns	11,181,861.31	0.0117	0.0002	49 B
Batch_Fan_In_Throughput	190.96 ns	3.747 ns	5.833 ns	5,236,621.08	0.0478	-	201 B

Note: Benchmarks ran without elevated priority (Permission denied for high priority). Results should still be representative for baseline comparison purposes.