Add Accept-Encoding header to fetch method

[gRegorLove]

Closes #208

[jtojnar]

As mentioned in #208 (comment), this option was renamed so ideally, it would be replaced with CURLOPT_ACCEPT_ENCODING conditional on version (see e.g. simplepie/simplepie#960).

This is not that critical right now since the current version of php keep the old name as alias. But will cause a warning once we introduce static analysis (something I plan to do soon).

[jtojnar]

Also this will break sites whose servers send encodings that curl does not support. We might receive bug reports like simplepie/simplepie#961 and then we will either have to decide that we refuse to support that, or implement retry like https://github.com/simplepie/simplepie/pull/962/commits

[Zegnat]

I do not think we have to be known for fetching, or supporting all those cases. So I would lean towards not supporting retries. It is always possible for someone to feed the data into Mf2\parse after they have gotten it themselves.

It would be good to support both constants, as @jtojnar says, based on the data that PHP gives us:

1. CURLOPT_ENCODING: “Available as of cURL 7.10 and deprecated as of cURL 7.21.6.”
2. CURLOPT_ACCEPT_ENCODING: “Available as of cURL 7.21.6.”

Probably just based on curl_version(), using version_compare()? Again, unlike SimplePie, I would not try to make that logic any harder than it has to be.