Procfile settings not read from mapped git branch

[jage]

Procfile is generated client side, uses settings from file-system instead of git branch

In Zambezi we have the ability to tweak the number of processes by setting a capistrano variable in production.rb and staging.rb. The variable is used to generate the Procfile, example: https://github.com/twingly/zambezi/blob/1eb92bebd863eee12c542c9c698624759188b588/config/deploy.rb#L9-L20

You can get some surprises by this though.

Since this is done client side, it does not use the settings checked in to the specific stage branch, instead it reads from the filesystem. If you run cap staging deploy when you're in the production branch, the production settings will be used. And vice versa.

This was first reported in #5