Docs: inference page confusions

[thegricean]

Comments apply to this page in the docs.

1. At the very outset: the comprehensive list of inference algorithms in one line (clickable, so it takes you down to the algorithm's description on the same page), followed by rules of thumb for using each and, if possible, a link to a model in forest that uses it.
2. For each inference algorithm, add a link to a description of that inference method (eg the ones on dippl), preferably to a description that also explains the role that the parameters play. For example, a description of MCMC would optimally explain the relation between number of samples taken, lag between samples, burning samples, etc. Or, for Enumeraction, the description would address why we would want different strategies/maxExecutions. If the description contained links to the literature on the topic (eg, Wikipedia page or the standard references), even better. Or, for Incremental MH, what is C3?
3. Rejection sampling: Can we use maxScore without incremental mode? Would be good to also give an example usage with incremental mode.
4. Kernels: it's nice that there are references included here for each kernel. It would be even cooler if there could be brief descriptions of the kernel either directly in the docs, or else if links to descriptions of the kernels could be included. What would be the most useful is if one could extract at a glance what the pros and cons of each kernel are, things to look out for, idiosyncracies, etc. "per-iteration" doesn't need the hyphen.
5. Incremental MH: give an example of query.add()
6. SMC: particle filter description is repeated #

[ngoodman]

I wonder if we should replace thunk with model throughout the Inference section?

[thegricean]

That would be great.

[chrisranderson]

I'll make the change that Noah mentioned.