Implement alternative WENO weights.

[ketch]

At ICOSAHOM, Debo Ghosh pointed out to me that recent improved versions of the WENO weights give much better results on general problems; see http://www.mcs.anl.gov/~ghosh/Files/2014_Ghosh_Baeder_JSC.pdf. I think we should implement the YC weights as default (leaving the JS weights, and possibly others, still as options).

Perhaps it makes more sense to implement these in PyWENO and then use it to generate code for SharpClaw.