Remove ColoredRecord class

[noamkush]

This intends to solve a rare problem I had encountered.
I use both colorlog and multiprocessing-logging. I noticed that in the subprocesses, if I call logger.exception, I do get the message but not the traceback. I tracked down the issue to this line, where multiprocessing-logging calls format(record), expecting the formatter to set record.exc_text. Since the ColoredFormatter wraps the record before calling the base class, the original record doesn't get it's exc_text set, and I lose my traceback.
I think that this issue is much easier to fix in colorlog. I removed the ColoredRecord and instead just set the escape code properties to the original record.

[borntyping]

In theory, accessing record.exc_text should to access the same attribute on the original record via __getattr__ on the ColouredRecord class (https://github.com/noamkush/python-colorlog/blob/eb6e2dca59c397586f5d5c3f48923da3d70a6127/colorlog/colorlog.py#L47-L48). Any idea why that doesn't work?

[noamkush]

Acessing record.exc_text would use __getattr__ but assigning a new value would use __setattr__.

[borntyping]

Ah, I misread the code you linked. I'd prefer to add __setattr__ if possible - my original intent with ColoredRecord is so that records don't get modified by colorlog (especially in setups with multiple handlers).

[noamkush]

Well, if you're also implementing __setattr__, you're modifying the record anyway.
I think the property names won't be used by other handlers.

[borntyping]

__setattr__ would only be used when another library modifies the record (i.e. it shouldn't pass through the __record attribute). I know the property names are very unlikely to be used elsewhere, but it'll be horrible to debug they do, such as a custom record class with a .reset() method.

[noamkush]

I tried to implement it with __setattr__ but it becomes an unworkable mess of __dict__ uses and infinite recursion. I think I'll just override the format method in my code.