HTML::DocumentFragment.parse should be able to accept an IO object

[flavorjones]

Describe the bug

HTML4::Document.parse, HTML5::Document.parse, and HTML5::DocumentFragment.parse all accept IO objects.

But HTML4::DocumentFragment.parse raises an unobvious exception when we do this.

To Reproduce

#! /usr/bin/env ruby
puts Nokogiri::HTML4::DocumentFragment.parse(StringIO.new("<div>hello</div>"))

emits:

/home/flavorjones/code/oss/nokogiri/lib/nokogiri/html4/document_fragment.rb:40:in `match?': no implicit conversion of StringIO into String (TypeError)

          path = if /^\s*?<body/i.match?(tags)
                                         ^^^^
        from /home/flavorjones/code/oss/nokogiri/lib/nokogiri/html4/document_fragment.rb:40:in `initialize'
        from /home/flavorjones/code/oss/nokogiri/lib/nokogiri/html4/document_fragment.rb:24:in `new'
        from /home/flavorjones/code/oss/nokogiri/lib/nokogiri/html4/document_fragment.rb:24:in `parse'
        from (irb):5:in `<main>'

Expected behavior

I'd like the HTML4::DocumentFragment.parse method(s) be able to handle an IO object, even if that means we just #read the whole string into a String.

[flavorjones]

@sharvy This might be a good first issue! There shouldn't be any C coding necessary, and it should be straightforward to write a test based on the reproduction script above.

I think the solution is basically something like what we have in the read_and_encode method in lib/nokogiri/html5.rb:

if tags.respond_to?(:read)
  tags = if encoding.nil?
    tags.read
  else
    tags.read(encoding: encoding)
  end
end

and I'd want to test both branches of that (with and without encoding specified).

[sharvy]

@flavorjones Thanks. Looks like a quick win. I'm on it.