Modernize TimeFrameLabeling

[shel-ho]

Addresses #79 and evaluates LID

[keighrim]

I understand @shel-ho is working on adding confusion matrix element to the report for this task. Can you work on "promoting" that feature to the common package? That'd be helpful to address #88 when we received the source assets.

[keighrim]

I think we need to get rid of --task argument, and instead replace it with a more generic argument to tell which label "to care". For example, instead of hard-coding the name "lid" and labels "en","es", or "lr", user should be able to pass something like --positive-labels en es at runtime and the script should take care of anonymizing all other labels to more generic labels like other0, other1, ... To generalize to future tasks.

[keighrim]

I updated the script with minor updates mentioned above (c6ccffc) . Still using --task and hard-coded names.
I re-ran the script on ambernet, SB preds and pushed new results under TF directory. @Brendayy I don't understand why you picked a completely arbitrary and new directory name to push the report the other day, but I deleted that commit. Please keep things organized.