*   `Estimator.export_savedmodel()` now includes all valid serving signatures
    that can be constructed from the Serving Input Receiver and all available
    ExportOutputs. For instance, a classifier may provide regression- and
    prediction-flavored outputs, in addition to the classification-flavored one.
    Building signatures from these allows TF Serving to honor requests using the

    // Gets the argument operands to the called function.
    operand_range getArgOperands() {
      // Merge `in_tensors` and `captured_tensors`. `operandSegmentSizes`
      // has two elements, so it will be merging two operand ranges (the two
      // variadic arguments). Attributes like `batch_timeout_micros` will not
      // make it to the return array.