Idea: Store self.log messages bodies of current_task, and add them to Crashed task body

[rakovskij-stanislav]

For better debugging purposes, I suggest to store all self.log message calls (even invisible on current verbosity level) and add them to Crashed task bodies on task failure. It will decrease the time required for troubleshooting.