Remove FPKM from results

[blahah]

R/FPKM are widely misused in practise and in the literature. It has been known since at least 2011 (RSEM paper) that they are not suitable for comparison between samples due to the library-specific normalisation factor. Unfortunately, most people use them that way.

TPM has all the benefits of R/FPKM but has the added benefit that the normalisation factor (1,000,000) is stable between samples. There is therefore no reason to use R/FPKM.

Salmon already promotes good practice in many ways, including reporting TPM. It should further promote good practise by not including FPKM in its results.

[rob-p]

In the latest release (along with other, even more-awesome new features), FPKM has been dropped from the output.