Juched78 log file fix

[rct]

• Moved @juched78's tornado log to file fix from a zip file attached to issue Logging in Tornado version doesn't write to file #36 to a branch/PR so the changes can be easily evaluated and/or merged.
• These changes are all from @juched78, this fix required some minor structural changes to complete configuration before setting up the logger.
• Since I didn't have a clean diff, the first commit adds @juched78's changes, which would have overwritten the plugin loading fix from Feature/tornado #42. The second commit fixes that. So this should now be a fairly clean patch other than the old bits that were commented out that should eventually be deleted.

[juggie]

Could this be cleaned up to not have commented out code all over the place.

[rct]

Will do. My intent was just to take @juched78's fixes out of the zip and turn into a diff/branch for testing/evaluation.

[juggie]

Looks much better but at quick glance I am not 100% sure why the logger.debug calls are being replaced with print() in 2 locations. I'll take a closer look shortly.

[rct]

Looks like he replaced those 2 lines because logger doesn't get imported until after config is done.

Though I just noticed that the check and exit if the config file didn't get loaded was removed.

[rct]

While this patch had removed the exit if config file not found, I don't think those lines worked. So I've added a commit to raise an error if the config file hasn't been processed. There are enough config defaults that things will sort of run without a config file. However it seems more appropriate, given the current setup to raise an error (which will force an exit).

[juggie]

We don't really need logger to know how to talk to the config module, we can easily have the logger be initialized in the main and tell it what kind of logging we want as part of application start up.

[rct]

I didn't really want to take ownership of this patch, but I did want logging to a file to work. My point of view on this PR at the moment, knowing very little about the Python logger, is that it is fairly reasonable at this point. Yes there have been 2-3 logger.debugs that have been replaced with prints because they are executed before the logger is configured.

I haven't had a chance to figure out what would be needed, if anything, to start up the logger in a default stderr config and then reconfigure it for file logging later. Sure it would be good for cleanliness, but is low on my priority list at the moment.