Class>>removeFromSystem silently skips removal if there are subclasses

[dalehenrich]

The original code was written years ago and it is likely that I duplicated the GBS logic - they have a similar restriction ...

There is no real harm in removing a class from Smalltalk when it has subclasses other than it is hard to find the class if you want it back for some reason:) - the system continues to function, only lookups will start failing ...

I raise this issue because "silently" failing to remove a class could be a source of Monticello load/unload issues and since I am opening up the [bugfix sweep pull request}(https://github.com/glassdb/pull/3) this might be a good time to put in some tests to a) prove that is can cause problems or b) see what havoc would be wreaked by making it an error or c) silently remove the requirement and move forward with life