Review Bot's logging was all over the place. Some code was using the

base logging module, some was calling logging.getLogger(), and some

was (rightfully) using Celery's get_logger() or get_task_logger()

functions. The Celery versions are pretty important, given how much

Celery controls the environment that our code runs in, so we actually

had some logging that didn't really work before.
This change aims to standardize logging, future-proofing it at the same

time. We now have our own reviewbot.utils.log.get_logger() function

that wraps Celery's get_logger()/get_task_logger(). It defaults to

creating task loggers, since most operations will run within the context

of a task.
BaseTool provides a new logger property for logging. It will

dynamically construct/retrieve a logger as needed, determining whether

that logger will be a task logger or a standard logger. This simplifies

tools even further, taking the guess work out of each implementation.
The log message formats have been altered to be a little bit nicer and

more consistent with typical log messages, and to include the logger

name (which Celery's default message format omits).
Some log levels have also been adjusted here and there, to cut down on

unwanted noise.

Unit tests pass for Python 2.7 and 3.x.
Verified that the logs on startup (for dependency checks and configuration)

were using standard loggers without task information, but that all the log

messages going through during a task run were showing task information.