Completion weighting

There are several systems that suggest completions at the moment it's just an ordered list

• Snippets
• API-injected completions
• .sublime-completions files
• Words in buffer

We can do better than this! Each of these categories should weigh their completion 0-1 on how confident they are this is relevant and then the list should be generated from that. This would make using completions much more tolerable.

There could be several weighting functions which then average the results.