Optimize Grape::Path -> Reduce Hash Allocation

[ericproulx]

This PR refactors the usage of Grape::Path and it comes with a lower Hash allocation. Most of the changes are in Grape::Path and the to_routes function in Grape::Endpoint

This is the result taken from the benchmark/compile_many_routes.rb

require 'memory_profiler'

class API < Grape::API
  prefix :api
  version 'v1', using: :path

  2000.times do |index|
    get "/test#{index}/" do
      'hello'
    end
  end
end

MemoryProfiler.report(allow_files: 'grape') do
  API.compile!
end.pretty_print

[dblock]

Alright thanks!