Expand acceptance tests to cover all of the scripts

[ctb]

Recent adventures suggest that we need better "real-world"/integration test coverage - particularly around the disk space checking, hashtable size checking, and threading. I propose the following:

• extract a subset of the 2013-khmer-counting/ pipeline into a k-mer counting protocol.
• extend the literate-resting approach into the metagenomics protocol (currently being done by @SherineAwad)
• write literate resting into some small recipes that can be used for lighter-weight testing as we discover points of fragility