* Document memory management approaches
* Change node args for docker to a better name (NODE_ARGS)
* Implement default node arg for docker `--max_old_space_size=512`
typeorm depends on ts-node as an *optional peer* dependency -- ts-node can be used in the typeorm cli to parse entities and run migrations.
However CM doesn't use typeorm CLI for running production so its not needed. And ts-node depends on typescript so npm install --production always installs both at about ~30MB.
I couldn't find a good way to remove peer deps ONLY for typeorm so instead just manually remove the folders from the prod install in the final layer of the docker.
* Switching to lsio as base means inclusion of s6-overlay for improved init, handling of default permissions, and a maintained base for security/bug fixes
* Also means we can use PUID and PGID as variables instead of requiring --user in run command (more friendly to unraid, portainer, etc...)
* BREAKING: mount directory for configuration,etc. has been changed to /config by default to align with lsio
* Clear npm cache after final production install to reclaim 100MB on image size
* More comprehensive dockerignore to prevent accidentally including development files
* Improve errors when checking if file is readable or directory/file is writable
* Implement custom winston-typeorm logging mappings so migrations are output at INFO level
* Add typeorm logging options to database config and default to displaying error, warning, and migrations
* Ensure directory is writeable before using log rotation transport for winston
* Move init logger into index so its always available
* Simplify printing stack for SimpleError when it's a cause
* Better handling for access permissions when reading operator config file
* Add hint to errors relating to permissions access if app detects its running in docker
* Change project location to /home/node to avoid root permissions when creating WORKDIR
* Create default config directory with correct permissions in the event no volume is mounted
* Set default db driver to better-sqlite3 since we control environment -- it has better performance and we know we can use a pre-built binary b/c of base docker image
* use --no-cache so repository index isn't included in image
* use FROM instructions to build project with dev dependencies, then copy built project to new stage and install production-only libraries
* properly ignore docs and node_modules with dockerignore
* Check for in-use web port before starting so we can display a useful error message
* Remove 'web' command and instead make it an optional argument for `run` (default to web)
* Update dockerfile to use run command
* Remove remaining commander option defaults since they are defined in config builder now
