Adds Dependency#version_project_id, and Project#dependent_projects_v2

[tiegz]

This denormalizes the Dependency#version#project association as Dependency#version_project, to speed up queries that do lookups for dependent projects, e.g. /npm/foo/dependents

Query using the existing @project.dependent_projects

SELECT  "projects".* FROM "projects"  
  INNER JOIN "versions" ON "projects"."id" = "versions"."project_id" 
  INNER JOIN "dependencies" ON "versions"."id" = "dependencies"."version_id" 
WHERE "dependencies"."project_id" = $1 
  GROUP BY projects.id 
  ORDER BY projects.rank DESC NULLS LAST;

Query using the new @project.dependent_projects_v2

SELECT DISTINCT "projects".* FROM "projects"  
  INNER JOIN "dependencies" ON "projects"."id" = "dependencies"."version_project_id" 
WHERE "dependencies"."project_id" = $1 
  ORDER BY projects.rank DESC NULLS LAST;

I'm not sure why the existing one uses a GROUP BY instead of DISTINCT, so please lmk if anyone has background on that.

Hard to say how much this will improve the query until it's deployed bc there are millions of dependency rows on production.

[tiegz]

My plan was to just deploy this, and if it helps, replace the other spots with this new method later.

[katzj]

I think this at least needs to be two PRs... one that has the schema change and one that uses it. The backfill is likely to take a while so we're going to need to have both in parallel for a bit

[tiegz]

Also considered dependent_project as a name? But this is more obvious at first glance.