[Backend][AIE] Add initial compilation support for Ryzen AI NPU/AIE

[chhzh123]

Description

This PR adds initial compilation support for Ryzen AI NPU/AIE, which contains the following parts:

1. MLIR-AIE codegen. We directly generate the source MLIR-AIE code (in MLIR) without calling the Python binding, which avoids environment conflicts. Currently, no automatic tile sizing is supported.
2. Ryzen AI NPU host codegen. Leverage the xrt_bo library.
3. End-to-end compilation flow from Python to AIE bitstream/execution.
4. Pure Python testing with Numpy. Input/output data are dumped into files. Also changed the Vitis HLS flow to directly read data from files instead of setting constant headers.

Examples

def _test_vector_scalar_add():
    Ty = int32
    M = 1024

    @df.kernel(mapping=[1])
    def core(A: Ty[M], B: Ty[M]):
        for i in range(M):
            B[i] = A[i] + 1

    top = df.build(core, target="aie")
    A = np.random.randint(0, 100, M).astype(np.int32)
    B = np.zeros(M).astype(np.int32)
    top(A, B)
    np.testing.assert_allclose(B, A + 1)
    print("PASSED!")

Checklist

• PR's title starts with a category (e.g. [Bugfix], [IR], [Builder], etc)
• Changes are complete (i.e. I finished coding on this PR)
• All changes have test coverage (It would be better to provide ~2 different test cases to test the robustness of your code)
• Code is well-documented