Handle failures in setup-wordpress.sh script more gracefully

[felixarntz]

Some of the longer tasks in setup-wordpress.sh can occasionally fail due to connectivity issues. Currently, this causes the entire process to fail, leaving the developer with an unusable environment (in most cases, any command from within the container would trigger an error message that the "getcwd() command could not be executed"). This problem became evident while testing at WordCamp Nordic (a common problem at WordCamps is the lack of a stable internet connection).

We already covered for such failures in the SVN checkout process. However, we should probably add similar error handling to other processes from that file. While most other tasks are required (while SVN checkout not always is), this still has significant benefits:

• By catching errors, we don't make the entire script fail. Apparently, having the script fail results in the environment being entirely broken.
• The setup-wordpress.sh script is run on start up. So even if something crucial fails, we could add a note and ask the developer to simply restart, which would cause the necessary tasks to be executed again. This is much easier (and quicker!) than completely destroying the environment and recreating it.
• You could even run the script manually again if you didn't want to restart.

What do you think @westonruter?

[westonruter]

One idea would be to incorporate something like travis_retry.

Something else that will help with performance is moving all of install-dependencies.sh into a Dockerfile, and then have Lando pull down the pre-build image so that none of the commands would be run, as I understand. This would also make a lando rebuild very fast.