Benchmark the platform dispatcher sequencers

The benchmark suite covers every production type that can run on Linux. The platform dispatcher sequencers are not covered, because each needs its own host:

• DispatcherSequencer (WPF), ControlSequencer (WinForms) and DispatcherQueueSequencer (WinUI), in both the plain and .Reactive packages. These need a Windows-targeted benchmark project.
• MauiDispatcherSequencer, in both packages. This needs a MAUI host.
• HandlerSequencer (Android) and NSRunloopSequencer (Apple), in ReactiveUI.Primitives and ReactiveUI.Primitives.Reactive. These need a device or simulator run.
• CoalescingDispatchState and DispatchWorkItem compile only into the Android and Apple targets, so they are reached through those sequencers.

Each sequencer should be measured the way the Blazor and Avalonia ones are in ReactiveUI.Primitives.Benchmarks: a burst of scheduled work items and an ObserveOn pipeline, against the platform's own dispatcher call and the System.Reactive scheduler for that host. Work must be drained deterministically rather than waited on.

Start in ReactiveUI.Primitives.Benchmarks by reading the existing Blazor and Avalonia benchmarks, then identify the Windows, MAUI, Android, and Apple benchmark hosts needed for the listed sequencers. Measure scheduled-work bursts and ObserveOn against each platform dispatcher and its System.Reactive scheduler, with deterministic draining. Done means all listed sequencers and platform-only dispatch types are covered on their target hosts.