[RUNTIME] Major code cleanup (#711)

This PR does the following:
- CUDA utilities (e.g., cuGetInfo) won't be compiled as part of libtriton.so anymore.
- Refactoring driver/llvm.cc to split it between PTX codegen and python.
- By extension this will also deprecate include/external so Triton won't have to live with a copy of some CUDA/Hip headers anymore.
- `triton-translate` becomes a `triton.tools.aot` Python utility that re-uses functions from the triton.compile sub-module.

test/Target/tritongpu_to_llvmir.mlir

@@ -1,4 +1,4 @@
-// RUN: triton-translate %s --target=llvmir | FileCheck %s
+// RUN: python3 -m triton.tools.aot %s --target=llvm-ir | FileCheck %s
 
 // == LLVM IR check begin ==
 // CHECK-LABEL: ; ModuleID = 'LLVMDialectModule'

test/Target/tritongpu_to_ptx.mlir

@@ -1,5 +1,4 @@
-// RUN: triton-translate %s --target=ptx --sm=80 --ptx-version=10000 | FileCheck %s
-
+// RUN: python3 -m triton.tools.aot %s --target=ptx --sm=80 --ptx-version=63 | FileCheck %s
 // CHECK-LABEL: // Generated by LLVM NVPTX Back-End
 // CHECK: .version 6.3
 // CHECK: .target sm_80