Add initial support for analyzing AMDGPU binaries (#900)

Add initial support for analyzing AMDGPU binaries (#900)

* remove unnecessary assertion for unknown phdr_type

* add initial stubs for amdgpu support

* add formatters for amdgpu

* add opcode table and register definition

* added support for register vector

* add untracked files

* commit stubs for xiaozhu to see

* initial cfg traversal done

* remove unused register-pair related files

* remove symbol patching since the 256 byte function header is no longer there

* Do not assert in getBaseRegister

* fix instruction length detection and decoding logic

* now decode will only decode operands for branch instructions

* added basic support for amdgpu in rose

* merged upstream

* fix a bug where we decode the vdst as ssrc

* improve the logic for computing len of instructions that might make use of literals, such that we no longer need to pass the buffer all over the place

* fix undefined behavior for std::transform

* Remove empty code blocks/Unreachable code/Unneeded
Comment out dead code
Restore sym_debug_parsing to default value 0

* Added meaningful name and comments to function prototype of getAMDGPURoseRegisters
Remove duplicate public modifiers in class declaration.

* remove unnecessary include of boost/format.hpp

* remove all amdgpu related stuff in dataflowAPI, and patched indirect analyzer so it skips for amdgpu

* renaming Arch_amdgpu to Arch_amdgpu_vega

* make changes to use namespace amdgpu_vega instead of amdgpu

* change the naming of c_NonReturnInsn to c_GPUKernelExitInsn as suggested

Co-authored-by: Xiaozhu Meng <mxz297@gmail.com>