ENH: execution could support rerunning failing notebooks/cells

[bsipocz]

Most of the notebook tutorials I maintain rely on remote data servers to run. Occasionally these produce timeouts and other flaky issues which are resolved after a restart. However, it would be super useful if I could avoid rerunning everything and just use a flag to ensure the problematic notebooks/cells are tried to be run N number of times (just as it works with pytest-rerunfailures)