DVCFileSystem.get() doesn't parallelize download

[grahameth]

Bug Report

Calling get on DVCFileSystem with multiple files doesn't parallelize the download.

Description

Reproduce

This example downloads multiple files with a single get call:

fs = DVCFileSystem(repo_url, rev=rev)
remote_files = [...]
local_files  = [...]
fs.get(remote_files, local_files, callback=cb)

It works, but it downloads 1 file at a time.

Expected

This example does the same, but accesses internal datastructures to make it faster:

fs = DVCFileSystem(repo_url, rev=rev)
remote_files = [...]
local_files  = [...]
fs._datafss[()].get(remote_files, local_files, callback=cb)

This internal get function parallelizes the download, and is thus about twice as fast on my machine.

Given that this functionality is already implemented, I find it a bit sad that it isn't used.

DVC version: 2.38.1 (conda)
---------------------------------
Platform: Python 3.9.15 on Windows-10-10.0.19045-SP0
Subprojects:
        dvc_data = 0.28.4
        dvc_objects = 0.14.0
        dvc_render = 0.0.15
        dvc_task = 0.1.8
        dvclive = 1.3.0
        scmrepo = 0.1.4
Supports:
        http (aiohttp = 3.8.3, aiohttp-retry = 2.8.3),
        https (aiohttp = 3.8.3, aiohttp-retry = 2.8.3),
        ssh (sshfs = 0.0.0)

[skshetry]

I'll suggest something like follows as a temporary workaround:

with Repo.open(repo_url, rev=rev) as repo:
	repo.dvcfs.get(...)

There are some inconsistencies between repo.dvcfs and DvcFileSystem, but for the purpose of fs.get(), they should work similarly (and be faster). They are still a bit inefficient, ideally we want to take the underlying filesystem of the remote (say s3fs) directly for better performance.