fix: fix the bug of considering the Path as a str when loading from TREC dataset (allenai#247)

Author: yzong12138

ir_datasets/formats/trec.py

@@ -159,7 +159,7 @@ def _docs_iter(self, path):
  elif path_suffix in ['.z', '.0z', '.1z', '.2z']:
  # unix "compress" command encoding
  unlzw3 = ir_datasets.lazy_libs.unlzw3()
- with io.BytesIO(unlzw3.unlzw(path)) as f:
+ with io.BytesIO(unlzw3.unlzw(Path(path))) as f:
  yield from self._parser(f)
  else:
  with open(path, 'rb') as f: