snake_case conversion eats existing underscores

[plinss]

This is a very useful plugin for converting legacy Python code to be PEP8 compliant.

However, a common convention in Python is to use leading underscores for private variables. When performing a snake_case conversion leading and trailing underscores are lost. e.g.:

_myVar -> my_var
myVar -> my_var

Ideally those should be preserved:
_myVar -> _my_var
myVar -> my_var

Additionally, existing multiple internal underscores are converted to single underscores. e.g.:
my__Var -> my_var