…specific

can re-use parts of this as new cases arise & refactor as more is learned. just
focusing on the specific at hand for right now, & in a way that won't require
public API changes in the future.

in expected usage, we'll do validation before passing algorithm name to Message.
so we should only be using default if client does not specify an algorithm.

if client specified an unknown algorithm, & we instead defaulted to sha256, that's
certain to result in a MIC mismatch & a failed transmission - but the silent
fallback will make the issue harder to troubleshoot. so better to raise & solve
the issue at the source.

examples:

  /Users/alexdean/.rvm/rubies/ruby-2.6.4/lib/ruby/2.6.0/net/protocol.rb:30: warning: method redefined; discarding old protocol_param
  /Users/alexdean/.rvm/gems/ruby-2.6.4/gems/net-protocol-0.2.1/lib/net/protocol.rb:32: warning: previous definition of protocol_param was here
  /Users/alexdean/.rvm/rubies/ruby-2.6.4/lib/ruby/2.6.0/net/protocol.rb:38: warning: method redefined; discarding old ssl_socket_connect
  /Users/alexdean/.rvm/gems/ruby-2.6.4/gems/net-protocol-0.2.1/lib/net/protocol.rb:40: warning: previous definition of ssl_socket_connect was here
  /Users/alexdean/.rvm/rubies/ruby-2.6.4/lib/ruby/2.6.0/net/protocol.rb:66: warning: already initialized constant Net::ProtocRetryError
  /Users/alexdean/.rvm/gems/ruby-2.6.4/gems/net-protocol-0.2.1/lib/net/protocol.rb:68: warning: previous definition of ProtocRetryError was here

  $ bundle exec rake test
  ... snip ...
  test/client_test.rb:160: warning: assigned but unused variable - file_name_received_by_bob
  test/client_test.rb:161: warning: assigned but unused variable - file_content_received_by_bob
  test/client_test.rb:177: warning: assigned but unused variable - file_name_received_by_bob
  test/client_test.rb:178: warning: assigned but unused variable - file_content_received_by_bob
  test/server_test.rb:31: warning: assigned but unused variable - status
  test/server_test.rb:66: warning: assigned but unused variable - status

This reverts commit 5b6d73a.

Will address this separately. See #21. At this point I'm not sure what the
actual problem or root cause are, so seems better not to mix this unrelated change
into the current feature addition in this PR.

leftovers from initial attempt to handle many different kinds of headers.
not needed now that this class is only focused on a single header type.