Add functions that suggests where to continue after invalid UTF-8/UTF-16 sequences

[tormol]

...To make it easy to continue if you're ignoring errors or inserting replacement characters.

This requires an error overhaul as many errors already include a position, and having it in two places would be dumb.
On the other hand, if we return (*Error,usize) the errors Display impl cannot use the usize.

Idea:struct SomeError<E>{err:E, len:usize}

[tormol]

For UTF-16 the next unit might always be (the start of) a valid codepoint.
... unless the error is that the second unit is a leading surrogate, and and the unit after it also is a leading surrogate; then ignoring the second unit would be more correct, but that seems unlikely.

[tormol]

I've decided to only advance one u8/u16 per returned error for predictability, but there could be helper functions that do this.