refactor: Extract evals to separate package (#2342)

* Remove tenacity dependency

- Retain experimental extra for both backwards compatibility and future use

* Remove unused retry configuration from BaseEvalModels

* refactor: Move token counting functionality off of `BaseEvalModel` (#2214)

* Standardize `model_name` parameter for `BaseEvalModels`

* Add abstract property to enforce `model_name` properties

* Factor out token counting functionality out of models

* Remove token counting functionality from models

* Update tests

* Update type annotations

* Add token counting module deprecation

* Restore missing `async_generate` method for LiteLLM model

* Update `generate` tests

* Add missing import

* refactor: rename BaseEvalModel to BaseModel (#2215)

* rename BaseEvalModel to BaseModel

* Update names

---------

Co-authored-by: Dustin Ngo <dustin@arize.com>

* Use private attribute for model name

* Move evals to packages

* Restore original evals

* Resolve imports

* Add dependencies

* Add tests

* Restore tenacity as a dependency

* Revert changes to experimental evals module

* Ruff 🐶

* Add test requirements

* Update import

* Fix imports, tests

* Resolve type-checker errors

* Add typing stubs

* Unnest utils, remove unused thread-based async

* Unnest functions

* Use absolute import

* Unnest templates

* Rename evaluator -> evaluators for consistency with new modules

---------

Co-authored-by: Xander Song <axiomofjoy@gmail.com>