osThread allocation failures should not abort the VM

Currently OSThread allocation will abort the VM if there is not enough memory. While the fact we are exhausting C-heap means we are likely to abort for one reason or another, it isn't the role of thread starting to act as that trigger. So `new OSThread()` becomes `new (std::nothrow) OSThread()` and the existing null checks actually serve a purpose.