Wish/to-do list for glmmADMB

* Can we add some sort of progress bar or "working ..." message in the absence of voluminous messages (difficult)
* Add more examples, tests, vignettes ...
* save (more) fitted models in inst/ so they can be used easily in examples without doing the fits each time
* simulate() method
* speed up (Skaug)
* more robust first pass, weighted SS or quasi-likelihood (Fournier)
* access to MCMC, parametric bootstrap tests (profiling???)
* MCMC fixes: access 'real beta'
* would making predictions into sdreport variables give (Wald) CIs that included uncertainty in Var-Corr?
* 

DONE:

* Switch to using R2ADMB as interface layer?? Compare R2ADMB and glmmADMB interface code, take the best of both ...
* work on predict() method
* truncated NB/NB1 models? 
 * (BMB,easy) provide compiled windows, macos binaries from www.math.mcmaster.ca in the absence of R-forge binaries?
 * (BMB) turn off "welcome to glmmADMB" banner (done)
 * allow alternative parameterization of NB?
 * (BMB,easy,*done*): get glmmADMB to save most output from ADMB unless a
verbose flag is set to TRUE (I think this stuff scares users)
 * (BMB, easy): implement a summary method that produces a coefficient
table in the 'standard' R format (coefficient, estimate, Z/t statistic
...) the big philosophical question would be whether to include a
two-tailed p value or not
 * (BMB, ?): add options for Pearson residuals?
 * (BMB, medium): implement a vcov() accessor method.  This would
require reading in an additional bit of information from the ADMB output.
 * (BMB,?): implement binomial models with size>1 (i.e., non-Bernoulli
models)
 * (BMB, ????): allow more flexible random effects specifications, etc.
