Normalize how objects are converted to text

[aaugustin]

One of the goals of the Debug Toolbar is to display a meaningful textual representation of a variety of objects.

However, depending on the objects, many things can go wrong. Not all objects implement __unicode__ or __str__. These methods may return a value of the wrong type or raise an exception. Attempting to display the exception may trigger an other exception or an infinite recursion.

(To be honest, the infinite recursion in __unicode__ or __str__ is the worst case I know.)

This ticket proposes to standardize how arbitrary objects are converted to text in an utility function and to use it throughout the project.

django.utils.encoding.force_text could be a good start; however, it's constrained by Django's backwards compatibility requirements and may not have all the features we desire.

[aaugustin]

This function, or a sister function, should also handle pretty-printing and indentation. Two issues suggest improving representation of objects in panels: #358 and #371.

[aaugustin]

<code>{{ value|pprint }}</code> in the templates seems to work quite well.

Non-ascii characters are escaped eg. \xe9 under Python 2, which is less readable, but not a bad compromise given the variety of crashing bugs we've had in the past.