Clean up code

[lukesonnet]

• Read through and simplify code
• Format according to Google's R style guide unless it is a big problem for backwards compatability
• Update documentation

[lukesonnet]

Finished auditing the R code. I didn't audit every function description or comment, but a good read through of the actual code. I also didn't format for style, for the most part. I'll do the cpp tomorrow.

All the changes worth noting, I think:

things that change usage

-Removed lambdaline and lambdapenalty options
-setting epsilon triggers truncation, truncation triggers default epsilon if not set. Also, warns if not truncating and n>=1000
-Removed sandwich argument from inference... if robust or clustered, then we will be sandwich estimator no matter what so robust and sandwich were redundant. We might want default to be sandwich instead of just inv hessian for KRLogit, but that's up to you
-Removed svd dependency and options to use it instead of RSpectra for eigen decompositions
-added an argument called lambdainterval, which sets the interval used in the optimize command, defaults to c(10^-12, 4)
-updated documentation on krls2

things that don't change usage

-Consolidated hyperctrl and control
-moved around some comments and some printing to make everything prettier, less 'debug' mode like