fixing retry protocol bug for not-completely-flushed requests

a bug was found where in certain cases, a server responds with an error
before the request fully buffers the body. Under retries, the request
is reset, however, the http/2 conn handler kept the last chunk around,
which it would flush before writing the second request body, resulting
in byte-accounting issues. Therefore, response clean up request state
before yielding.

lib/httpx/connection/http2.rb

@@ -226,6 +226,10 @@ module HTTPX
     end
 
     def on_stream_close(stream, request, error)
+      log(level: 2) { "#{stream.id}: closing stream" }
+      @drains.delete(request)
+      @streams.delete(request)
+
       if error && error != :no_error
         ex = Error.new(stream.id, error)
         ex.set_backtrace(caller)
@@ -241,9 +245,6 @@ module HTTPX
           emit(:response, request, response)
         end
       end
-      log(level: 2) { "#{stream.id}: closing stream" }
-
-      @streams.delete(request)
       send(@pending.shift) unless @pending.empty?
       return unless @streams.empty? && exhausted?