Drop excluded projects from Jira links, mark partial untagged counts, find the highest tag on cut lists

[lesnik512]

Three follow-ups to --exclude and untagged rows. The report schema stays at 5.

Jira links to excluded projects. Each Jira issue's GitLab links still named projects matched by --exclude, and the skill reads those links to say which services an issue touches. Links to excluded projects are now dropped when they are read, in --jira runs and for the Jira issues found on rows in --group/--project runs. for_issues builds its projects from the same links, so its separate filter is gone.

Untagged count as a lower bound. When the walk stops at RELEASE_SCOPE_MAX_COMMITS and no row in the range has a tag, more untagged rows may lie below the cut. The page now shows N+ untagged and "At least N ..." in that case, in English and Russian. This is computed on the page from truncated and untagged.rows, with no new report field.

Next tag when the tag list is cut short. With more tags than the paged list holds, the highest version could be missing, so next_tag came out too low. When the list is cut, one more request asks GitLab for tags with order_by=version&sort=desc (documented in doc/api/tags.md), and those names count toward the highest version. Without a cut, there is no extra request.

Each fix started from a failing test, except the page change: the page's JavaScript has no tests, so I checked it in a browser against a sample report. Only a cut range without a tag got the +; a cut range with a tag and an uncut range kept the exact count.