Evaluate system file simplified API

[neubig]

Starts to address #31

This simplifies the API to evaluate systems to be a single API call such as the following:

result: dict = client.evaluate_system_file(
    system_output_file="sst2-output.txt",
    system_output_file_type="text",
    task="text-classification",
    system_name="test_cli",
    metric_names=["Accuracy"],
    source_language="en",
    target_language="en",
    dataset="sst2",
    split="test",
    shared_users=["[email protected]"],
)

[pfliu-nlp]

I think these changes in this PR will not break the implementation of the one in the teaching class. （but let me know if it will affect so that I can make corresponding revisions)

[neubig]

I don't think it will, but I'd actually suggest that you set a specific version that you use for the class to prevent any future changes from causing issues. That's what I did as well: https://github.com/neubig/minbert-assignment/blob/main/setup.sh#L13

[pfliu-nlp]

"suggest that you set a specific version that you use for the class to prevent any future changes from causing issues"

yeah, this makes sense