Add seek step to enable faster silence detection

[marciomazza]

Silence detection can be very slow on large files.
One way of speeding it up is to seek for silence chunks at steps, instead of trying every single millisecond as a chunk start.

This pull adds such a seek step to the detection, as a parameter at the end of the involved functions, with a default value of one. That makes it backwards compatible.

One possible future improvement would be to parallelize the detection.

[marciomazza]

I noticed the max columns used now in the project is 99, by running this at the project root:

find -name '*.py' | xargs cat | awk '{print length}' | sort | tail -n 1

This config makes autopep8 respect that maximum line length.