HTTP response to request for asset contains in some cases header transfer-encoding chunked

Steps to reproduce

1. Upload a pdf
2. run 

   curl -svko /dev/null -X GET 'http://localhost:8080/magnoliaPublic/dam/jcr:518d21bb-b021-43a4-b56d-fe13c16ed0e1/Kube.pdf' --http1.1 -H 'Range: bytes=0-20971519'

    (use bigger number then size of the document here, or 8192, see the comment)

3. check output, when using http1.1 you can see the transfer-encoding set and content-length is 0, when using http1.0 content-length is 0

Expected results

no transfer-encoding header just content-length to the right size

.. Justify non-trivial expectations with a link to a doc or a relevant discussion.

Actual results

Workaround

disable filter for pdf's

Development notes