Limit jobs and add pagination

[marcosvm]

Limits jobs and adds pagination for #49

• add pagination for jobs in the data layer, with default of 10 jobs
• add limit of 50 jobs
• add pagination in the GraphQL layer

Supports basic pagination for jobs adding the query arguments page and size, with queries like:

  jobs(page:3, size: 2)  {    
  
    id    
    repoSlug
    log
  }
}

The data layer will limit the number of jobs to 50 and the default page size is 10.

[coveralls]

Coverage increased (+0.2%) to 83.412% when pulling cc56d9e on marcosvm:marcosvm/limit_jobs into b601645 on elixir-bench:master.

[AndrewDryga]

Cursor based pagination is probably not the best option if we allow going deep into the list. Project is public and google would nail our database by indexing last pages: https://use-the-index-luke.com/no-offset. @tallysmartins what do you think?

[tallysmartins]

Hi, thanks for your feedback @AndrewDryga

I didn't think about this problems regarding cursor based pagination and I didn't get how it avoids google crawling, for both solutions can't we just set up the robots.txt to limit the page indexing?

And the offset-less approach seems to be a better implementation indeed. @marcosvm can you take a look at the article given in the comment from @AndrewDryga ?

[marcosvm]

@tallysmartins Sure thing. I'm looking it at it.

[AndrewDryga]

There is a package that uses id + timestamp cursors for paging: https://github.com/duffelhq/paginator.

@tallysmartins cursor pagination should not solve issues with crawling, but a core issue why they could nail our DB. Crawlers would index all pages over and over again, following very deep links where people won't ever look (eg. indexing results from 10000 builds ago). If we decide use offset based paging queries to respond to crawlers can be very expensive slowing down an entire system.

This is not very important while we are super small though.