Add frozen_string_literal everywhere and fix string mutation with force_encoding

[AnotherJoSmith]

This PR has two commits:

1. Add the magic comment # frozen_string_literal: true to every file. This will help with migration to ruby 3, where all literals will be frozen. This functionality was added in ruby 2.3.
2. Duplicate strings before calling force_encoding in Request. String#force_encoding is a mutating function, which will raise a RuntimeError if the response body is a frozen string.

This is mostly causing issues for us with testing using stub_response in our application, since we have enabled frozen string literals.

For example, the following test fails when frozen_string_literals are enabled:

httparty/spec/httparty/request_spec.rb

Lines 551 to 559 in da1b1ad

	it "should assume utf-16 little endian if options has been chosen" do
	@request.options[:assume_utf16_is_big_endian] = false
	response = stub_response "C\x00o\x00n\x00t\x00e\x00n\x00t\x00"
	response.initialize_http_header("Content-Type" => "text/plain;charset = utf-16")
	resp = @request.perform
	expect(response_charset).to_not be_empty
	expect(resp.body.encoding).to eq(Encoding.find("UTF-16LE"))
	end

The issue also exists with this gem's tests, which you can see by running git checkout 9c4c7a0 && bundle exec rake.

This should not be a breaking change and is being done by other gems as well (e.g. stripe/stripe-ruby#649).

@jnunemaker If you're not interested adding the frozen_string_literal comment everywhere, I'd be happy to pull out 1f29818 as a separate PR as well.