#expiring_url returns public_url when not using AWS to prevent exceptions when Fog Local is swapped for Fog AWS in development.

This was previously fixed in pull request #539 but no spec was added at that time, and it has since been broken.

I fixed it and added a spec.

Reverting this for now. This commit would better be in a minor release,
since it break everyone's application. Going to reopen the ticket and
mark it to be merged in next minor release.

This reverts commit b7b2b843.

* Fixed tests and functionality for new fog and removed the fixed version from the gemspec
* Set time to 3600 from the current time so expiring actually works

* Added test for checking if path gets interpolated correctly without passing
  :path or :url options
* gsub default :rails_root filesystem-like path to nothing for Fog storage,
  just like in S3 Storage.

`s3_protocol` now returns the protocol without a colon. If you need a
colon, you can pass in `true` as a second argument.

This might not be a best solution, so any better patch that does not
introduce a regression is welcome.

Fixes #921

* Add StyleTokenNotFound in Errors
* Adjust tests to make sure :style present where not currently defined

Fixes #902

Fixes #883

Re-fix of #838, #766, #911

IMPORTANT: There is a bug in the current release of Fog (1.3.1) that will
return a broken expiring URL when the bucket name is valid for a subdomain.
This is fixed per https://github.com/fog/fog/pull/857 but not released
in the gem yet. I have added currently FAILING tests for these cases that
should pass when a new Fog is released.

Make sure that we close opened files after we're done with them.

Needs work for S3 Attachments.

Fixes #738

This adds functionality to add additional params to the querystring for
S3 expiring URLs. The reason for this is if you want to override
response_content_type or response_content_disposition with expiring
URLs, you have to change Amazon's signature, which gets signed before
adding additional options to the querystring. Because it's added later,
and because Amazon's signature includes the full URL, the signature is
bad and the request fails.

To use this feature:

    has_attached_file :avatar,
                      :s3_url_options => { :response_content_disposition => "inline" }

Additionally, you can pass a lambda and it'll be evaluated when the URL
is generated. If, for example, the content type is incorrect in Amazon
(either empty or application/octet-stream), you can effectively cast the
response from Amazon as a particular content type.

    has_attached_file :avatar,
                      :s3_url_options => lambda {|model| { :response_content_type => model.avatar_content_type } }

Fixes #743

Fixes #703