Update rstudio-bnlearn-cpu with new packages

[rbartelme]

• [x] Add installs for stan, rstan, tidybayes, ggfortify, brms
• [x] Rebuild container and tag as cyversevice/rstudio-bayes-cpu:3.6.2
• [x] Local test of build
• [x] push to cyverse dockerhub
• [x] update tool configs in VICE
• [x] test on VICE

[rbartelme]

Note: iRODS v4.1 is not currently working, but this is non-essential for this specific analysis.

[rbartelme]

Stan is a real pain to install inside a container, this is going to be quite complex.

[rbartelme]

https://github.com/stan-dev/rstan/wiki/Installing-RStan-on-Linux and https://github.com/antoine-sachet/rocker-ML/ helped get Stan running inside the rocker/verse container environment locally. Testing the container on VICE today.

[rbartelme]

rstan version 2.19 causes R session to abort on the DE when it loads. It does not abort when the container is run locally. It's possible this is a server side K8s issue.

[rbartelme]

Container runs on VICE but dies when I run brm function, same abort message that the DE throws when I simply try to load the package.

[rbartelme]

After talking to Julian, Sarah and @tyson-swetnam about this, it seems the clang configs were incorrect for the user rstudio that runs the container. Local test of the container was able to run the diagnostic script for rstan:

stancode <- 'data {real y_mean;} parameters {real y;} model {y ~ normal(y_mean,1);}'
mod <- stan_model(model_code = stancode)

Pushing to CyVerse VICE, will test with the diagnostic script there. If the diagnostic script runs without throwing C++ errors, then I will likely need to just refactor my model for the canopy_height.

[rbartelme]

DE is still throwing errors, but the container runs fine locally. Sarah is looking into it.

[rbartelme]

Works fine on the DE after Sarah recompile rstan in the container with the K8s admin tools. Container runs fine on GPU06 and is updated to R4.0 with Ubuntu 18 LTS.