Cannot use tensorrt.broadcast with int8 inputs

[pranavm-nvidia]

The following program (after disabling casting workarounds in full):

print(tp.ones(shape=(2, 2), dtype=tp.int8))

IRs:

==== Trace IR ====
def main() -> (
    %t3 : tensor<?x?xi8:gpu:0>
):
    %t1 = constant(shape=(), dtype=float32, device=gpu:0) : tensor<f32:gpu:0>
    %t2 = cast(%t1 : tensor<f32:gpu:0>, dtype=int8) : tensor<i8:gpu:0>
    %t0 = constant(shape=(2,), dtype=int32, device=gpu:0) : tensor<2xi32:gpu:0>
    %t3 = broadcast(%t2 : tensor<i8:gpu:0>, %t0 : tensor<2xi32:gpu:0>) : tensor<?x?xi8:gpu:0>
    return %t3

==== MLIR ====
module @"outs_%t3_0" {
  func.func @main() -> tensor<?x?xi8> {
    %cst_f32 = tensorrt.constant dense<1.000000e+00> : tensor<f32>
    %0 = tensorrt.cast %cst_f32 : tensor<f32> to tensor<i8>
    %cst_i32 = tensorrt.constant dense<2> : tensor<2xi32>
    %1 = tensorrt.broadcast %0 broadcast_dims<> shape(%cst_i32 : tensor<2xi32>) : tensor<i8> to tensor<?x?xi8>
    return %1 : tensor<?x?xi8>
  }
}

Fails with:

MTRTException: failed to run pass pipeline
    [optimizer.cpp::getFormatRequirements::3300] Error Code 2: Internal Error (Assertion !n->candidateRequirements.empty() failed. No supported formats for [tensorrt.slice] (%t3) , please check the TensorRT operator doc (https://docs.nvidia.com/deeplearning/tensorrt/operators/docs/index.html) for supported formats.)
    (%t3) error: failed to translate function 'tensorrt_cluster' to a TensorRT engine