Bump LLVM and StableHLO (#2598)

Bump LLVM to `5e5a22caf88ac1ccfa8dc5720295fdeba0ad9372` and StableHLO to
`83f095e7217c897f1eccac5652600ceb944cb0e0`.

Bazel GHA:
https://github.com/sjain-stanford/torch-mlir/actions/runs/7027647674

externals/llvm-project

@@ -1 +1 @@
-Subproject commit 28b27c1b10ae8d1f5b4fb9df691e8cf0da9be3f6
+Subproject commit 5e5a22caf88ac1ccfa8dc5720295fdeba0ad9372

externals/stablehlo

@@ -1 +1 @@
-Subproject commit 77a59815a82b34f7b08ed2d42a711d9920682d0e
+Subproject commit 83f095e7217c897f1eccac5652600ceb944cb0e0

test/Conversion/TorchConversionToMLProgram/basic.mlir

@@ -10,7 +10,7 @@
 // CHECK:           %[[NEXT_SEED:.*]] = arith.addi %[[MUL]], %[[INC]] : i64
 // CHECK:           %[[INSERTED:.*]] = tensor.insert %[[NEXT_SEED]] into %[[GLOBAL]][] : tensor<i64>
 // CHECK:           ml_program.global_store @global_seed = %[[INSERTED]] : tensor<i64>
-// CHECK:           return %2 : i64
+// CHECK:           return %[[NEXT_SEED]] : i64
 module {
   func.func @f() -> i64 {
     %seed = torch_c.get_next_seed : () -> i64

utils/bazel/WORKSPACE.bazel

@@ -33,6 +33,10 @@ llvm_configure(
     },
     targets = [
         "X86",
+        # The bazel dependency graph for mlir-opt fails to load (at the analysis step) without the NVPTX
+        # target in this list, because mlir/test:TestGPU depends on the //llvm:NVPTXCodeGen target,
+        # which is not defined unless this is included.
+        "NVPTX",
     ],
 )