Refactor default judges

[SecroLoL]

Default judges used to be created with the syntax

from judgeval.scorers import JudgmentScorer
from judgeval.constants import APIScorer

scorer = JudgmentScorer(threshold=0.5, score_type=APIScorer.FAITHFULNESS
...

We can now create them by directly importing

from judgeval.scorers import FaithfulnessScorer

scorer = FaithfulnessScorer(threshold=0.5)

This is a stylistic change but also reveals strong future benefit as we add scorers that require specific args, such as the new JSONCorrectness scorer, which needs a schema field

scorer = JSONCorrectnessScorer(schema=...)

We wouldn't have been able to provide this with the previous schema, but now we can.

Major changes

1. Implement each default judgment scorer as its own class
2. Modify functions that run evaluations (standard evals and span-level evaluations) to accommodate this
3. Update all existing UTs to be compatible, + add UTs for all default scorers

[JCamyre]

Take a look at my comments, this is sick

[JCamyre]

I will handle this, I'll share my thoughts in Slack. In my multi-step eval PR, I added a project and trace name, which will tie in nicely to generate automatic eval run names (to improve UX while remaining clear).

[SecroLoL]

Nice B)