dvcfs: implement get_file

Currently we use `open`, which is much slower. For example,

```
$ dvc get dvc-bench data/small/dataset
```

goes down from 110sec to 50sec.

Discovered as a part of https://github.com/iterative/dvc-bench/issues/419

dvc/fs/dvc.py

@@ -361,6 +361,19 @@ class _DVCFileSystem(AbstractFileSystem):  # pylint:disable=abstract-method
         info["name"] = path
         return info
 
+    def get_file(self, rpath, lpath, **kwargs):  # pylint: disable=arguments-differ
+        key = self._get_key_from_relative(rpath)
+        fs_path = self._from_key(key)
+        try:
+            return self.repo.fs.get_file(fs_path, lpath, **kwargs)
+        except FileNotFoundError:
+            _, dvc_fs, subkey = self._get_subrepo_info(key)
+            if not dvc_fs:
+                raise
+
+        dvc_path = _get_dvc_path(dvc_fs, subkey)
+        return dvc_fs.get_file(dvc_path, lpath, **kwargs)
+
 
 class DVCFileSystem(FileSystem):
     protocol = "local"

pyproject.toml

@@ -35,7 +35,7 @@ dependencies = [
     "configobj>=5.0.6",
     "distro>=1.3",
     "dpath<3,>=2.0.2",
-    "dvc-data>=0.40.4,<0.41",
+    "dvc-data>=0.41.0,<0.42",
     "dvc-http",
     "dvc-render>=0.1.2",
     "dvc-studio-client>=0.1.1",
@@ -57,7 +57,7 @@ dependencies = [
     "requests>=2.22",
     "rich>=12",
     "ruamel.yaml>=0.17.11",
-    "scmrepo>=0.1.11,<1",
+    "scmrepo>=0.1.12,<1",
     "shortuuid>=0.5",
     "shtab<2,>=1.3.4",
     "tabulate>=0.8.7",