* make robust to multiple ADMB runs in the same directory: locking?
* add example to help file!
* attach data as well when checking bounds in DATA section
* don't try to check data/params at all when check is set to "ignore"
* fix integer trap! (somehow)
* document write_files; modularize further (allow turning off compile, etc.)
* weaken/eliminate ggplot2 dependency
* comparisons with PBSadmb?
* implement "fixed" argument (set phase <0)
* support for transformations (on the fly)?
* add MCMC quantiles etc. to summary: coeftab?
* confint method?
-   "quad" if only se/vcov known
-   "profile" if known
-   MCMC: quantile/HPDinterval if known
* better setup detection, error trapping:
   comments on adcomp/adlink/admb scripts
* more checking of dat/pin files:
* allow for more sections in TPL file to passed through "as is"
* test/document examples of random effects in vignette:
* ?? formula interface???
* read bar (binary) rather than par file?
* read hessian file?
* evaluate objective function by fixing vals
* restructure into three chunks: 
    (1) check and construct TPL/data files; (2) run ADMB;
    (3) pull results into R (and clean up?) -- 
       generate shell script for run?  generate
       versioned TPL file?  
     (all to simplify use e.g. by people who will upload to a cluster)

=======

* add phase support

* finish bounds support: check bounds on initial parameters?

* better strategy for saving originals when changing case/auto-generating
inputs -- temporary filenames, or temp directory??

* TEST check for variables with "." in the name: stop? change to "_" and warn?      
* check for random effects vectors (don't redefine if already there)
* parameter/data order checking? (done, I think)
* more checks/flags of compiling step -- stop if error
* more info on thinning etc. of MCMC saved/printed in final object
* allow just compilation, just run from previous compilation ...
* double-check parnames in read_pars(), compare with Nielsen version
