Allow multiline tensor definitions in rank-profile

[kjocevicius]

Is your feature request related to a problem? Please describe.
Currently it's not allowed to split tensor declaration into multiple lines. We're implementing reranking, using ONNX model in rank-profile. It accepts multiple tensors as inputs, and it is very inconvenient to have them in single line. While example is only 8 features, we have a tensor with 20 features as well.

function x_categorical() {
    expression: tensor<int8>(d0[1], d1[1], d2[8]):[attribute(first_id), attribute(second_id), attribute(third_id), attribute(fourth_id), attribute(fifth_id), attribute(sixth_id), attribute(seventh_id), some_other_parameter]
}

Describe the solution you'd like
It should be allowed to split tensor declaration to multiple lines, like:

function x_categorical() {
    expression: tensor<int8>(d0[1], d1[1], d2[8]):[
        attribute(first_id), 
        attribute(second_id), 
        attribute(third_id), 
        attribute(fourth_id), 
        attribute(fifth_id), 
        attribute(sixth_id), 
        attribute(seventh_id), 
        some_other_parameter
    ]
}

Describe alternatives you've considered

• Making some template to work around this
• Using Python to describe ranking profile, but it might be only long term solution to rewrite everything later