Render buffers with 32-bit float depth

A render buffer created with type QRhiRenderBuffer::DepthStencil seems to use 24-bit integer depth. It is sometimes useful to have render buffers with 32-bit float depth.
OpenGL lets you do that with
glRenderbufferStorage(GL_RENDERBUFFER, GL_DEPTH32F_STENCIL8, ...);
Why not let the user select the image format of a render buffer in RHI? The current API with just two types seems a bit limited.