Misc bug fixes + Allow sleuth_prep to process just one sample

[warrenmcg]

Hi @pimentel,

First of all, I noticed to small errors in my contributions to sleuth_prep that will result in problems for users:

• A typo caused an error if factor levels were supposed to be dropped.
• If there is an error in one of the later mclapply jobs, but the first job is successful, then the code will continue as if everything is ok and this may result in a cryptic downstream error for the user. This updated code checks all of the jobs to make sure they were all successful, and throws an error if at least one of them failed.

To expand on pull_request #92, I went ahead and updated the code to prevent errors due to a one-column data.frame being converted to a vector using drop = FALSE. I also added a warning if the user supplies one sample but also a model, since the modeling aspects of sleuth don't work when you only have one sample. I have confirmed this new code works by running the small_test_data Kallisto file, which I'm incorporating into some new tests (separate pull request).

[pimentel]

right on, thanks yet again!

I've been thinking about allowing 1 sample analysis for a while but have wanted to put safeguards in. I think the warning is sufficient for now. we will probably expand it later.

best,

harold