add dilation and fix add ops in pytorch emitter

[garry1ng]

Problem 1: I try to convert a caffe layer to pytorch which is an elewise layer containing multiple bottom layers. The pytorch emitter only handles two-element add ops.

layer {
bottom: "fc8_1"
bottom: "fc8_2"
bottom: "fc8_3"
bottom: "fc8_4"
top: "fc8"
name: "fc8"
type: "Eltwise"
eltwise_param {
operation: SUM
}
}

Problem 2: The pytorch emitter doesn't support dilation features though they exist in IR code.

Platform (like ubuntu 16.04/win10): ubuntu 16.04

Python version: Python 3.6

Source framework with version (like Tensorflow 1.4.1 with GPU): caffe GPU 1.0

Destination framework with version (like CNTK 2.3 with GPU): pytorch 1.2 with GPU