Running ilp selector after max cov can produce wrong results

[Chadi-akel]

If you run the ilp selector after max-cov, results may be wrong.
Indeed, max cov selector can for example marks a region as being too small if its self coverage is
under a limit (1% default value). If you then run the ilp selector, this region will still be considered as too small because this information is written in graph_.pkl file.

A workaround is to always run cere profile before running a selector as it will reset the graph (but keep everything else, dumps, replays etc...)