spec: wrap/override fsspec httpfs

[pmrowla]

Fixes iterative/dvc-objects#195

[pmrowla]

It would probably be better for our filesystems to just wrap fsspec directly instead of using the fs.fs redirection that we have right now, but this fixes the issue until we can actually get around to refactoring dvc_objects.fs

[skshetry]

It would probably be better for our filesystems to just wrap fsspec directly instead of using the fs.fs redirection that we have right now, but this fixes the issue until we can actually get around to refactoring dvc_objects.fs

Ideally, we should just directly use fsspec-provided filesystems. There should be no wrappers.

[skshetry]

@pmrowla, I have approved the PR, I have a few minor suggestions inlined. But I think it'd be better to upstream this in fsspec's http implementation itself.

Also, I'll encourage you to open a discussion regarding batching issue in fsspec, as this dvc-http is not the only place this might break.
We also have similar code in dvc-azure, but I am not sure if that breaks anything or not.

https://github.com/iterative/dvc-azure/blob/78e3ff570d176bec954915847c403122a2b8cf95/dvc_azure/__init__.py#L178-L180