changelog creation script does not structure the output sensibly

the current implementation of the changelog script is very naive. when the one who runs it fails to edit the output into something sensible prior to inviting further participation, frustration among maintainers and a substandard quality of the final changelogs is almost guaranteed.

here are some specific items that need addressing:

• the easy part: tags should be considered case-insensitively to avoid pointlessly duplicated sections.
• the harder part: the script needs to impose a sensible hierarchy. there are two major parts to that:
  • the component hierarchy needs to be standardized.
    • instead of accepting arbitrary tags, only specific ones should be considered
    • there should be a tree with a predefined structure. in qtbase, the 5.6.0 changelog is a good example, while the 5.9.3 one is a bad one ("changes" just isn't a reasonable section title in a file named "changes").
      • the script should ignore the provided tag order and instead impose the predefined hierarchy on the entries
  • tags which don't denote components need to be dealt with properly. specifically, it's currently completely unclear what to do with "important behavior changes" and "platform specific changes", which leads to a basically random structure. imo,
    • the "important behavior changes" tag should divide the main contents into two top-level sections, so every component potentially appears twice in the file. all component-specific hierarchy should appear below the main sections.
    • attempts to group by platform should be dropped entirely. instead, each entry should be tagged with a platform as necessary.
      • we are about to emphasize this structure in jira as well.