Add python bindings

[j2kun]

In-progress work adding python bindings to HEIR

[j2kun]

On 2024-01-26 Maksim Levental and I were exploring this on a livestream and found that the python binding support for bazel in upstream MLIR was hard-coded to tensorflow's configuration decisions (see llvm/llvm-project#79676 and https://github.com/j2kun/test_mlir_bazel_pybind). That resulted in the upstream fix to the bazel build rules getting merged and so we can proceed with adding HEIR-specific python bindings here.

[j2kun]

After another session with @makslevental, we got to the point in these two related PRs

• Fix mlir bazel overlay makslevental/llvm-project#3
• hacks j2kun/heir#8

Where we can build the python bindings, but then trying to run the hello world script results in the following error at https://github.com/llvm/llvm-project/blob/5fc28e7a8d989fffafec2301529d0164a28799b2/mlir/include/mlir/IR/StorageUniquerSupport.h#L161

LLVM ERROR: Registering an interface for an attribute/type that is not itself registered.

Putting print statements at that location to show the list of registered dialects and types, it shows only the builtin dialect types have been registered

REGISTERED Types:
builtin.builtin.tuple
builtin.builtin.f32
builtin.builtin.memref
builtin.builtin.f128
builtin.builtin.unranked_memref
builtin.builtin.none
builtin.builtin.function
builtin.builtin.unranked_tensor
builtin.builtin.vector
builtin.builtin.opaque
builtin.builtin.tensor
builtin.builtin.index
builtin.builtin.integer
builtin.builtin.f64
builtin.builtin.f80
builtin.builtin.bf16
builtin.builtin.f16
builtin.builtin.f8E4M3FN
builtin.builtin.f8E5M2FNUZ
builtin.builtin.f8E4M3FNUZ
builtin.builtin.complex
builtin.builtin.f8E5M2
builtin.builtin.tf32
builtin.builtin.f8E4M3B11FNUZ

However, the registration code in InitAllDialects runs before this without error.