Refactor download (#4384)

* Refactor download

* Fix textvqa dataset

* Move download module one level up

* Fix test mock fsspec

* Fix docs

* Fix import of DownloadConfig

* Refactor mock_download_manager

* Refactor DownloadConfig

* Fix docs

* Add sections to docs of builder classes

* Use relative imports

* Fix path

* Fix import

Commit from https://github.com/huggingface/datasets/commit/82258b47dcf6e12a9b9f2aec68ea18fb3fd3d9fc

americas_nli.py

@@ -18,7 +18,6 @@
 import csv
 
 import datasets
-from datasets.utils.download_manager import DownloadManager
 
 
 _CITATION = """
@@ -133,7 +132,7 @@ class AmericasNLI(datasets.GeneratorBasedBuilder):
             citation=_CITATION,
         )
 
-    def _split_generators(self, dl_manager: DownloadManager):
+    def _split_generators(self, dl_manager):
         dl_paths = dl_manager.download(
             {
                 "dev_data": _DEV_DATA_URL,