Use registered multipart/form-data parser

[moberegger]

With the changes made in #472, openapi_first will now always use RequestBodyParsers::MultipartBodyParser to parse a multipart/form-data request. This PR makes it so that it will use the registered parser (which by default is RequestBodyParsers::MultipartBodyParser).

This is to support any custom 'multipart/form-data' parsers that may have been registered. For example, with

OpenapiFirst::RequestBodyParsers.register(
  'multipart/form-data',
  MyCustomMultipartBodyParser,
)

openapi_first will now use MyCustomMultipartBodyParser.

[ahx]

Hi. Thanks for the PR. This makes the whole story about request body parsers more coherent.

This PR makes it apparent, that I got a bit lazy in #472 and vibe coded the thing, because I wanted to get the issue closed without looking more into finding a better interface that uses the same code path for content with "encoding" or without "encoding".

[ahx]

Maybe we should change the interface for parsers to always pass an arbitrary options hash like

class MyCustomMultipartBodyParser
  def initialize(options={})
  end
end

OpenapiFirst::RequestBodyParsers.register(
  'multipart/form-data',
  MyCustomMultipartBodyParser,
)

# … and internally it always calls `.new` on the class
RequestBodyParsers[content_type].new({encoding: …})

This would mean an interface change from parser_class.new.call(request) to parser_class.new(options_hash).call(request).