revert transform_keys change

[JamesChevalier]

Closes #271 by reverting the transform_keys change

omniauth ultimately defines options as a Hashie::Mash not a Hash:
https://github.com/omniauth/omniauth/blob/master/lib/omniauth/strategy.rb#L547
https://github.com/omniauth/omniauth/blob/master/lib/omniauth/key_store.rb#L5

While Hashie::Mash does implement transform_keys, it does not behave the same way as the Hash implementation does:

Hashie::Mash.new(one: "1", two: "2").transform_keys(&:to_sym)
{"one"=>"1", "two"=>"2"}

Hashie::Mash.new(one: "1", two: "2").to_hash.transform_keys(&:to_sym)
=> {:one=>"1", :two=>"2"}

---

🤔 An alternative could be to rewrite the existing line to:

@options = @@default_options.merge(options.to_hash.transform_keys(&:to_sym)) 

Never mind this idea 😬

Ah! Another option is to add `.with_indifferent_access` when `@@default_options` is created: https://github.com/oauth-xx/oauth-ruby/blob/62d3f3e52daa565a04e508df915456a21a7faf37/lib/oauth/consumer.rb#L33-L73

irb(main):001:0> hh = {one: '1', two: '2'}
=> {:one=>"1", :two=>"2"}
irb(main):002:0> hm = Hashie::Mash.new(one: "one", two: "two")
=> {"one"=>"1", "two"=>"2"}
irb(main):003:0> mm = hh.merge(hm)
=> {:one=>"1", :two=>"2", "one"=>"one", "two"=>"two"}
irb(main):004:0> hh = {one: '1', two: '2'}.with_indifferent_access
=> {"one"=>"1", "two"=>"2"}
irb(main):005:0> mm = hh.merge(hm)
=> {"one"=>"one", "two"=>"two"}
irb(main):006:0> mm[:one]
=> "one"

I don't expect the removal of Hashie from omniauth to be possible, based on how purposeful they were in switching to it:
https://github.com/omniauth/omniauth/wiki/OmniAuth-1.0#changes-to-the-auth-hash

[JamesChevalier]

Closing in favor of #273
Sorry for the noise

Update: Re-opening because 273 inadvertently added a dependency on ActiveSupport

[pboling]

In the OAuth2 gem we actually decided to solve this by using the rash_alt gem, which is a subclass of Hashie::Mash, and it is working great there. I think perhaps we could do the same thing here.

Note: faraday had the same idea!

[JamesChevalier]

maintainers went with #276