Change the line parsing and storage approach to reduce the overall line processing time

[JoeKar]

The intention came up out of the discussion for #3127.
We realized that the line is processed multiple times as bytes to receive the resulting runes and their count.

This can be significantly reduced by storing already the decoded runes per line while parsing that line the first time. Afterwards this line doesn't need to be decoded again and the count can be accessed by a simple usage of len().

Since this is a huge rework and change of one of the basics it might brake one or more interfaces used by plugins.

The change is still ongoing and a lot of depending functionalities need to be changed. So please expect rebases from time to time.
I created the PR anyway since a lot of very important discussion was already ongoing within #3127 which shall be continued in here to keep the former one clean now.

In case someone else likes to support here then please provide PRs to performance/rune-lines as base branch and we can apply them.

BTW:
make test runs successfully already.

This PR will continue #3149.

[JoeKar]

@dmaluka
You should now be able to work directly at performance/rune-lines, in case you find some time to do so. 😉
A lot of work is still present to exchange these LineBytes() and util.CharacterCount() accesses + the stuff I missed.