fix: Extending TS can now create lower cost alignment without panic

[jeeeesper]

Apart from some reformatting, the only change is that the assertions on the new cost being equal are removed.

Closes #89

[sebschmi]

Nice, thanks! I propose that we replace the former assertions assert_eq!(new_cost, initial_cost); with initial_cost = new_cost, to avoid increasing the cost again, as described in the issue.

[jeeeesper]

Hey @sebschmi , this should be fine and all tests pass, but maybe you wanna take a look before we merge. Also, I've done a bit of refactoring to avoid panics in general, to allow for more graceful failure.

[sebschmi]

Thanks, looks very nice. Two of the comments you made indicate a bug to me, as I commented at the specific lines. This kind of bug is something we may actually run into, in case an alignment starts with a TS, or in case the TS inner is at the boundary of the sequence.

[sebschmi]

Thanks! I went over things again, and noticed that we would now panic in some cases in debug mode, but allow them in release mode. We (I) never run this code in debug mode though, because it is too slow. So best to still either panic when the assert indicates a programming error within the function, or return a Result::Err in case of violated preconditions of the method. But in case of violated preconditions, we can also just document that false is returned. But then again we may not want to panic in debug mode then...

I think cleanly using either asserts or return Result::Err is the best way to deal with these.

I marked some of the spots of this happening in the first method, but since the others are similar I didn't mark them.

[jeeeesper]

I've implemented it now so that false is returned on error since the docs imply that anyway, and an error is logged in that case.

[jeeeesper]

And now it's even rebased 🫡