perf: collapse 4-layer async forwarding chain in test execution

[thomhurst]

Summary

The per-test entry point chains through 4 async layers before any real work happens, each materializing its own state machine. Estimated 2 state machines per test removed.

Evidence

Trace (inclusive %):

11.54% TestRunner.ExecuteTestAsync
11.28% TestRunner.ExecuteTestInternalAsync
11.10% TestCoordinator.ExecuteTestAsync           ← pure forward
11.10% TestCoordinator.ExecuteTestInternalAsync
10.13% TestCoordinator.ExecuteTestLifecycleAsync
 9.94% TestExecutor.ExecuteAsync

TUnit.Engine/Services/TestExecution/TestCoordinator.cs:50-51 is a pure forwarding async method:

public ValueTask ExecuteTestAsync(AbstractExecutableTest test, CancellationToken cancellationToken)
    => ExecuteTestInternalAsync(test, cancellationToken);

ExecuteTestLifecycleAsync is an inline-able wrapper on the no-retry branch.

Proposed fix

• Inline pure-forward TestCoordinator.ExecuteTestAsync into its caller.
• Inline ExecuteTestLifecycleAsync into ExecuteTestInternalAsync for the no-retry path.
• Audit TestRunner.ExecuteTestAsync → ExecuteTestInternalAsync for the same pattern.

Expected impact

• 2-3 async state machines per test removed.
• Directly reduces AsyncMethodBuilderCore.Start 2.93% self time observed in the trace.

Files

• TUnit.Engine/Services/TestExecution/TestCoordinator.cs:50-51, 53, 279
• TUnit.Engine/Scheduling/TestRunner.cs:48, 91, 125
• TUnit.Engine/TestExecutor.cs:69