Benchmarks introduction does setup wrong

[AMindToThink]

The introduction says:
from deepeval.benchmarks.task import MMLUTask

but really, it is

from deepeval.benchmarks.mmlu.task import MMLUTask

[AMindToThink]

https://docs.confident-ai.com/docs/benchmarks-mmlu

The steps described result in the model generating long outputs, not single answers.
Also, batching doesn't work because MMLU has prompts of different lengths.

[penguine-ip]

Hey @AMindToThink can you please create a quick PR for this?

[AMindToThink]

I made a pull request to change one location with the wrong MMLUTask import. There may be others.

And I am unable to fix the other issues.