Make it easier to recover from a bad install

Today I helped a few members with a broken GDK. There were a number of issues:

1. The port file had invalid data: it had something like 3000 echo 3001>port. This caused make to fail right out of the box, and nothing was set up properly. Perhaps earlier validation should be done here with a better warning message. #659 (closed)
2. Even after fixing the port file, make didn't check out all the dependencies (e.g. gitlab-shell, gitaly, etc.). I had to run make gitlab-shell-setup and make gitaly-shell-setup etc. to get this working. #629 (closed)
3. ruby was using v2.4.4, not v2.6.3, for some reason, so one of the startup Ruby scripts died because it was using 2.6.3 syntax. I didn't see a .ruby-version in the GDK dir though. I had to force a rvm use 2.6.3 to get this to work. Perhaps we need to do a version check to make sure it's set up properly. Closed by #628 (closed)
4. Even after having all the dependencies, gdk update failed because the gitlabhq_development didn't exist. I had to run gdk run gitaly db first in one window, and then bundle exec rake dev:setup to seed the database. Perhaps we need to detect the database doesn't exist and reinitialize it. #667 (closed)
5. Building the elasticsearch-indexer failed due to the icu4c issue linked in the troubleshooting. I had to set export PKG_CONFIG_PATH="/usr/local/opt/icu4c/lib/pkgconfig:$PKG_CONFIG_PATH". #631 (closed)
6. Jaeger conflicted with both the CE and EE ports. Perhaps we should automatically use different ports for an EE installation. (Irrelevant once we have the single codebase). #632 (closed)
7. One person wasn't aware that you had to run bundle install; yarn in the gitlab directory after updating the gitlab/ dir. Some people use gdk update, but not always. #633 (closed)