CURLOPT_ACCEPT_ENCODING set to NULL

[carlopi]

Relevant documentation at https://curl.se/libcurl/c/CURLOPT_ACCEPT_ENCODING.html

I am considering whether having an option (that can be set user side), would be best here in case of regressions.

Fix was proposed by @ywelsch in the context of wider testing they performed where files uploaded to S3 like:

aws s3 cp file.json.gz s3://<bucket>/file.json.gz --content-type gzip

would be unzipped twice since information is not propagated between layers.

[carlopi]

Relevant in the context of duckdb/duckdb#20977

[rustyconover]

Hi, this PR has caused some problems with me and I just opened an issue to re-enable cURL's support for Content-Encoding. With this change no responses are compressed with gzip/deflate and we're using more bandwidth than we need.

But even worse, if I want to handle the decompression myself in an extension, now when the server sends Content-Encoding: zstd (because I added an Accepts-Encoding header), curl rejects the response with CURLE_BAD_CONTENT_ENCODING and the entire request fails.

[rustyconover]

PR is #278