Extra examples from Musketeer work

[ms705]

This pull request brings in some extra Naiad example jobs developed at Cambridge for the Musketeer paper (dl.acm.org/citation.cfm?id=2741968):

• Netflix movie recommendation (dense matrix version, two implementations).
• PageRank (two implementations, derived from GraphLINQ examples).
• Single-source shortest path (SSSP).
• k-means clustering.
• JOIN and PROJECT micro-benchmarks.
• TPC-H query 17 (Lindi and enhanced Lindi versions).
• Top-shopper workflow.

The implementations are hand-written Naiad baselines, not auto-generated jobs from Musketeer. We used these jobs as our baselines in the paper.

We may need to do some code cleanup and add better documentation in order to merge these, but I thought it'd be best to open a PR and start the discussion, so that we can track changes here.

[michaelisard]

I am fine with adding these examples, but it might make sense to merge them with the fault tolerance branch which is really the active development branch now (e.g. it includes the improved graphlinq, as well as the fault tolerance work) instead of release_0.5. We should figure out what to do so that it builds with Mono without removing the Windows-only projects too.

[ms705]

Good points. I'll remove the changes that break the Windows build and rebase the PR onto the development branch. The only argument in favour of merging this into release_0.5 is that the new examples would become available in the stable branch even before the 0.6 release is cut from the development branch. I'm happy to go either way :)