use FunctionalOperator::clean to have consistent results in wccl-run (variables will not roll over to subsequent operator runs)

change -I switch in wccl-run and wccl-rules to compose a list of corpora paths instead ofacting as a stdin input switch. Use -I - for stdin.

refactor wcclrun a bit, add formatting options: --tabs makes it output plain tab separated data, -l and -g control local / global token counts