Feature/lambada evaluator

[bmosaicml]

What does this PR do?

This PR implements a composer-native evaluator for Lambada.

What issue(s) does this change relate to?

Previously we were dependent on the Eleuther eval harness to run Eval, which was implemented inefficiently and couldn't support multi GPU. We now support LAMBADA with multi-gpu and enjoy linear scaling w/ num GPUs.

This is related to the following JIRA https://mosaicml.atlassian.net/browse/RESEARCH-345

Before submitting

• Have you read the contributor guidelines?
• Is this change a documentation change or typo fix? If so, skip the rest of this checklist.
• Was this change discussed/approved in a GitHub issue first? It is much more likely to be merged if so.
• Did you update any related docs and document your change?
• Did you update any related tests and add any new tests related to your change? (see testing)
• Did you run the tests locally to make sure they pass?
• Did you run pre-commit on your change? (see the pre-commit section of prerequisites)

[dakinggg]

Left a bunch of comments, mostly about integrating everything nicely with composer, and other clean code stuff like tests and docstrings. The implementation itself looks correct to me. Happy to help with any of the comments related to integrating it with composer if you want!