[libc] Use intermediate float operations on the GPU (llvm#155869)

Summary:
Move the GPU config to the common directory instead of copying it and
enable the intermediate float option in the math library.

Author: jhuber6

libc/config/gpu/amdgpu/config.json libc/config/gpu/config.jsonlibc/config/gpu/amdgpu/config.json renamed to libc/config/gpu/config.json

@@ -39,7 +39,7 @@
   },
   "math": {
     "LIBC_CONF_MATH_OPTIMIZATIONS": {
-      "value": "(LIBC_MATH_SKIP_ACCURATE_PASS | LIBC_MATH_SMALL_TABLES | LIBC_MATH_NO_ERRNO | LIBC_MATH_NO_EXCEPT)"
+      "value": "(LIBC_MATH_SKIP_ACCURATE_PASS | LIBC_MATH_INTERMEDIATE_COMP_IN_FLOAT | LIBC_MATH_SMALL_TABLES | LIBC_MATH_NO_ERRNO | LIBC_MATH_NO_EXCEPT)"
     }
   }
 }