extract --continue, optimise for interrupted big files

#7526 added `borg extract --continue` with a very simple and uncritical change, but it does not help if extraction is interrupted in the middle of a single huge file: incomplete files are deleted and extracted from scratch.

#1665 is a more complex and more critical PR, but likely could solve this issue (if the remaining issues in that PR are solved).

Start with the existing `borg extract --continue` behavior from #7526 and review the more complex approach in #1665, including its remaining issues. Done means interrupted extraction of a single huge file can continue without deleting and re-extracting the incomplete file.

Written by the indexing model from the issue text.