Patch: bugs in sleuth_results & other miscellaneous fixes

Hi @pimentel,

`sleuth_results` bugs:

I discovered two bugs in sleuth_results: 1) setting show_all = F still results in all target_ids being returned. 2) setting pvalue_aggregate = T results in the extra annotation data being missing.

I made the following changes to the function to hopefully correct this behavior.

`spread_abundance_by` speed-up

I also earlier had made a faster version of spread_abundance_by, and realized that it make it in the propagation forward of merged changes. That version is found below, and I have confirmed that it produces identical results in at least three separate datasets with different target_ids. Yet, it has a 3-fold increase in speed. It's a small gain overall, but it greatly affects the sleuth_live experience since plotting functions heavily use spread_abundance_by, and it affects to a small extent sleuth-ALR overall speed.

NAMESPACE issue with `transform_status`

When running devtools::document(), the NAMESPACE updated transform_status from a regular method to an S3method.

sample_to_covariates `data.table` error, addressing issue #153

When a sample_to_covariates table is provided to sleuth_prep as a data.table object, it passes the data frame check in place, but causes downstream problems. I put in a line to coerce any table that passes the check to a data frame.

Best, Warren

pachterlab / sleuth