Adjust IP after an AV during an STEP with FEATURE_EMULATE_SINGLESTEP enabled

[thaystg]

Fix step in a line that will throw an AV

Without this PR in this function IsSafeToHandleHardwareException this other function ExecutionManager::IsManagedCode(controlPc) was returning FALSE, basically because we were trying to find the patched IP in the Jitted code and then it wasn't found, right now I'm adjusting the patched IP to the original IP before trying to check the is it's in a managed code.

Fixes #95189

[noahfalk]

Rather than adding a new handler, could we merge the fixup logic into the g_safeExceptionCheckFunction? It looks like other fixups are already being applied in that function:

runtime/src/coreclr/vm/exceptionhandling.cpp

Line 5433 in 8839704

// Pretend we were executing the barrier function at its original location

Not required, just figured it might cut down needing to setup and define new callbacks.

[noahfalk]

This function returns TRUE/FALSE, but the callsite ignores the value. Any reason to have it if we aren't using it?

[noahfalk]

Any reason why we only invoke pThread->HandleSingleStep() in some cases and not others? Could we do something like this in IsSafeToHandleHardwareException() and remove the HandleSingleStep call from HandleHardwareException?

#ifdef FEATURE_EMULATE_SINGLESTEP
Thread *pThread = GetThreadNULLOk();
if (pThread != NULL && g_pDebugInterface != NULL)
{
    HandleSingleStep(contextRecord, exceptionRecord, pThread);
}
#endif

It feels easier to understand if we always invoke it from one place rather than conditionally invoke it from two different place.

[thaystg]

@tommcdon the tests are passing.
@noahfalk I think I addressed all your comments.

[noahfalk]

The updated version of the change looks like it still calls HandleSingleStep from two different places, once here and once over in HandleHardwareException.

I was suggesting that we converge those into a single call here. Is there a reason we need two different calls? If yes it would be great to add a comment saying why we have both, and if not it would be nice to get rid of the HandleHardwareException portion and do it all here.