[draft] build: update llvm tag to 14212c96

[ashay]

Do not review yet, waiting for CI to pass.

Summary of changes:

• Updated MHLO commit to build with LLVM commit hash 14212c9
• Re-enabled two failing TOSA tests, thus resolving Github Issue Investigate failure of ElementwiseReluModule and ResNet18StaticModule TOSA tests #1231
• Replace deprecated hasValue() with has_value()
  (https://reviews.llvm.org/D131349)

[ashay]

@fortianyou Any idea why the MHLO tests failing? Here's a portion of the stack trace from the assertion failure:

#0  __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:50
#1  0x00007ffff7c64537 in __GI_abort () at abort.c:79
#2  0x00007ffff7c6440f in __assert_fail_base (fmt=0x7ffff7dcd128 "%s%s%s:%u: %s%sAssertion `%s' failed.\n%n", assertion=0x7fffc05ed7e0 "hasVal",
    file=0x7fffc11bec87 "torch-mlir/externals/llvm-project/llvm/include/llvm/ADT/Optional.h", line=215, function=<optimized out>)
    at assert.c:92
#3  0x00007ffff7c73662 in __GI___assert_fail (assertion=0x7fffc05ed7e0 "hasVal",
    file=0x7fffc11bec87 "torch-mlir/externals/llvm-project/llvm/include/llvm/ADT/Optional.h", line=215,
    function=0x7fffc09f6a1d "T &llvm::optional_detail::OptionalStorage<mlir::DenseElementsAttr, true>::value() & [T = mlir::DenseElementsAttr]")
    at assert.c:101
#4  0x00007fffc552b09b in mlir::mhlo::(anonymous namespace)::ConvolutionOpGeneralConversion::matchAndRewrite(mlir::mhlo::ConvolutionOp, mlir::mhlo::ConvolutionOpAdaptor, mlir::ConversionPatternRewriter&) const ()
   from torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir/_mlir_libs/libTorchMLIRAggregateCAPI.so
#5  0x00007fffc552053b in mlir::OpConversionPattern<mlir::mhlo::ConvolutionOp>::matchAndRewrite(mlir::Operation*, llvm::ArrayRef<mlir::Value>, mlir::ConversionPatternRewriter&) const ()
   from torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir/_mlir_libs/libTorchMLIRAggregateCAPI.so

The error can be reproduced by running python3 -m e2e_testing.torchscript.main --config=mhlo -v -s, although the individual test python3 -m e2e_testing.torchscript.main --config=mhlo -v --filter=AdaptiveAvgPool2dUnitOutputSizeStaticModule_basic passes.

[ashay]

Closing, since I will attempt a new LLVM merge tomorrow (Monday) morning.