[HLSL] Implement `Load(int, out uint)` overloads for buffer types for DirectX

[bogner]

All of the HLSL buffer (and texture) types have overloads of Load like RWBuffer::Load(int, uint). We need to implement these in HLSLExternalSemaSource where we have the loads without the status bit.

These functions should call the appropriate buffer load intrinsic directly rather than emitting getpointer like the other loads do, since that's the only way we'll be able to map the status bit to the load of the right location. We also need to consider SPIR-V, which is in this separate issue: #166954