Handle duplicate joins

[matthewmcgarvey]

Fixes #370

Summary

Querying with joins can sometimes produce invalid sql. The two cases that this PR handles are:

1. When they potential join is to the table being queried
2. When the potential join has already been joined

Description

These situations are rather easy to get to when users attempt to make composable methods that limit the scope of queries. Say an app has the concept of a Business and a User and if the user belongs to a business they are considered an Employee of that business. If the app allowed users to access the api, it might make sense to limit their API access to businesses they are an employee of. Rather than having to remember how to scope their access correctly every time you have a new query, you could make helper methods like (granted this is convoluted, I'm sure there are more realistic ways to get in this situation):

def user_scoped(query)
  scoped_user_query = UserQuery.new.id(current_user.id)
  query.where_users(scoped_user_query)
end

def business_scoped(query)
  scoped_business_query = user_scoped(BusinessQuery.new.id(current_business.id))
  query.where_businesses(scoped_business_query)
end

(Those methods would require that Business has a has many through association to User through Employee)
The problem arises when you want to provide an endpoint that returns all the employees for the user's current business. The action would do something like:

employees = business_scoped(EmployeeQuery.new)

You end up with an error: table name "employees" specified more than once (PQ::PQError) because the BusinessQuery adds a join to limit by the user id and the main query is on the employees table.

Note

I'm sure there are situations this doesn't handle and this might even cause issues so PLEASE poke holes in this change.

[jwoertink]

Looks good! Just needs a rebase or pull from master. After everything passes feel free to merge 👍