refactor: deploy torchserve using v1 protocol instead of v2 protocol

Although pytorch supports KServe v2 protocol, its implementation doesn't fully adhere to the v2 protocol and models that output more than one array cause the KServeEvelope to crash. More at: pytorch/serve#3405

Author: gabrielmscampos