Flexeval

What's Changed
* Debug `BLEU` by ryokan0123 in https://github.com/sbintuitions/flexeval/pull/18
* Update tests for metrics by ryokan0123 in https://github.com/sbintuitions/flexeval/pull/19
* Deal with a case when stop_seq is tokenized into an empty string in `HuggingFaceLM` by ryokan0123 in https://github.com/sbintuitions/flexeval/pull/20


**Full Changelog**: https://github.com/sbintuitions/flexeval/compare/v0.4.0...v0.4.1

What's Changed
* Implement `max_instances` option by ryokan0123 in https://github.com/sbintuitions/flexeval/pull/13
* `flexeva_lm` の `EvalSetup` 用のインターフェイスを更新 by ryokan0123 in https://github.com/sbintuitions/flexeval/pull/12
* Refactor `flexeval_file` and `flexeval_pairwise` by ryokan0123 in https://github.com/sbintuitions/flexeval/pull/14
* Use `loguru` for loggers by ryokan0123 in https://github.com/sbintuitions/flexeval/pull/15
* Remove unused code by ryokan0123 in https://github.com/sbintuitions/flexeval/pull/16
* rename `OpenAIChatGPT` to `OpenAIChatAPI` by ryokan0123 in https://github.com/sbintuitions/flexeval/pull/17


**Full Changelog**: https://github.com/sbintuitions/flexeval/compare/v0.3.3...v0.4.0