[AMDGPU] Update amdgpu_cs_chain_preserve docs. NFC

We no longer allow calls to functions with the `amdgpu_gfx` calling
convention from functions with the `amdgpu_cs_chain_preserve` calling
convention. See D153517.

Also mention that we can't have a chain call from
amdgpu_cs_chain_preserve using more VGPRs than it has received.

Differential Revision: https://reviews.llvm.org/D156408

Author: rovka

llvm/docs/AMDGPUUsage.rst

@@ -1196,6 +1196,8 @@ The AMDGPU backend supports the following calling conventions:
                                      uniform control flow.
 
      ``amdgpu_cs_chain_preserve``    Same as ``amdgpu_cs_chain``, but active lanes for VGPRs starting at v8 are preserved.
+                                     Calls to ``amdgpu_gfx`` functions are not allowed, and any calls to ``llvm.amdgcn.cs.chain``
+                                     must not pass more VGPR arguments than the caller's VGPR function parameters.
 
      ``amdgpu_es``                   Used for AMDPAL shader stage before geometry shader if geometry is in
                                      use. So either the domain (= tessellation evaluation) shader if