[libc] Rework the RPC interface to accept runtime wave sizes (#80914)

Summary:
The RPC interface needs to handle an entire warp or wavefront at once.
This is currently done by using a compile time constant indicating the
size of the buffer, which right now defaults to some value on the client
(GPU) side. However, there are currently attempts to move the `libc`
library to a single IR build. This is problematic as the size of the
wave fronts changes between ISAs on AMDGPU. The builitin
`__builtin_amdgcn_wavefrontsize()` will return the appropriate value,
but it is only known at runtime now.

In order to support this, this patch restructures the packet. Now
instead of having an array of arrays, we simply have a large array of
buffers and slice it according to the runtime value if we don't know it
ahead of time. This also somewhat has the advantage of making the buffer
contiguous within a page now that the header has been moved out of it.

Author: jhuber6

libc/src/__support/GPU/amdgpu/utils.h

@@ -17,9 +17,6 @@
 namespace LIBC_NAMESPACE {
 namespace gpu {
 
-/// The number of threads that execute in lock-step in a lane.
-constexpr const uint64_t LANE_SIZE = __AMDGCN_WAVEFRONT_SIZE;
-
 /// Type aliases to the address spaces used by the AMDGPU backend.
 template <typename T> using Private = [[clang::opencl_private]] T;
 template <typename T> using Constant = [[clang::opencl_constant]] T;
@@ -108,8 +105,11 @@ LIBC_INLINE uint64_t get_thread_id() {
          get_num_threads_x() * get_num_threads_y() * get_thread_id_z();
 }
 
-/// Returns the size of an AMD wavefront. Either 32 or 64 depending on hardware.
-LIBC_INLINE uint32_t get_lane_size() { return LANE_SIZE; }
+/// Returns the size of an AMD wavefront, either 32 or 64 depending on hardware
+/// and compilation options.
+LIBC_INLINE uint32_t get_lane_size() {
+  return __builtin_amdgcn_wavefrontsize();
+}
 
 /// Returns the id of the thread inside of an AMD wavefront executing together.
 [[clang::convergent]] LIBC_INLINE uint32_t get_lane_id() {

libc/src/__support/GPU/generic/utils.h

@@ -16,8 +16,6 @@
 namespace LIBC_NAMESPACE {
 namespace gpu {
 
-constexpr const uint64_t LANE_SIZE = 1;
-
 template <typename T> using Private = T;
 template <typename T> using Constant = T;
 template <typename T> using Shared = T;
@@ -55,7 +53,7 @@ LIBC_INLINE uint32_t get_thread_id_z() { return 0; }
 
 LIBC_INLINE uint64_t get_thread_id() { return 0; }
 
-LIBC_INLINE uint32_t get_lane_size() { return LANE_SIZE; }
+LIBC_INLINE uint32_t get_lane_size() { return 1; }
 
 LIBC_INLINE uint32_t get_lane_id() { return 0; }
 

libc/src/__support/GPU/nvptx/utils.h

@@ -16,9 +16,6 @@
 namespace LIBC_NAMESPACE {
 namespace gpu {
 
-/// The number of threads that execute in lock-step in a warp.
-constexpr const uint64_t LANE_SIZE = 32;
-
 /// Type aliases to the address spaces used by the NVPTX backend.
 template <typename T> using Private = [[clang::opencl_private]] T;
 template <typename T> using Constant = [[clang::opencl_constant]] T;
@@ -95,8 +92,8 @@ LIBC_INLINE uint64_t get_thread_id() {
          get_num_threads_x() * get_num_threads_y() * get_thread_id_z();
 }
 
-/// Returns the size of a CUDA warp.
-LIBC_INLINE uint32_t get_lane_size() { return LANE_SIZE; }
+/// Returns the size of a CUDA warp, always 32 on NVIDIA hardware.
+LIBC_INLINE uint32_t get_lane_size() { return 32; }
 
 /// Returns the id of the thread inside of a CUDA warp executing together.
 [[clang::convergent]] LIBC_INLINE uint32_t get_lane_id() {