Add --iterate option

[sourcefrog]

Adds a --iterate option that will skip mutants that passed in a previous run. The idea is that if you are adding tests that, presumably, only improve coverage, you can easily find out if you've now fixed every previous gap.

Fixes #53

This will only be stable if you're using the same filter options (--file) etc on consecutive runs.

Fuzzy matching?

There is a question whether it's better to match the line and column numbers, or better to consider mutants as matched even if they've moved a bit. It's easy to imagine that in the process of improving coverage, some code would move around a bit and ideally we'd detect that it moved but did not change. On the other hand, ignoring the line and column is likely to be ambiguous when there are similarly-named mutants within the same file or function, which might cause some things to be skipped incorrectly.

The process of catching mutants is probably often about adding more test coverage. Since that occurs typically in either integration test files or in unit tests at the bottom of files, it typically shouldn't cause code to move? If the code under test needs to be changed to improve coverage, perhaps it's better to re-test it. So, overall, I think I won't do any fuzzy matching at this point.

TODO

• Add an --iterate command-line option (it would not make sense to have in the config.)
• Read in caught.txt and previously_caught.txt before creating the new output directory.
• Write all of them into previously_caught.txt in the new output directory.
• Skip previously caught mutants.
• Integration type tests.
• Document how to use this.
• Document the new file in mutants.out.
• Test that --list excludes previously-caught mutants, but does not update the files.
• Test there's no previously_caught.txt without --iterate.