Add metrics for user snapd version per snap

As snapd continues to release new features to users over time, certain users are held back in their snapd versions. Whilst many distributions are up relatively up to date with snapd packages, there are some that aren't, and the re-exec functionality of snapd isn't always available or always used (typically due to users not being aware it exists or of its significance).

Some features might be only available in newer snapd clients and packagers may wish to prioritise testing and documentation around snapd versions that are still heavily in use. Whilst the `assumes` functionality exists, packagers may wish to avoid using it to avoid cutting off installations on older snapd versions and it's hard to understand just how many that is by looking at the existing metrics (I.E, OS distribution does not correlate 1:1 to snapd version, and there's simply a lot of supported distros). In some other cases it might be clear that it's worth avoiding newer functionality at all due to low benefit vs the amount of older clients it may cut off or otherwise adversely affect. It would be easier for a packager to make an informed decision with actual statistics available.

Ultimately, I expect rather than have a nice gradient where the latest versions of snapd heavily dominate and the older versions of snapd are all but obsolete, you would instead see signficant peaks at various points (E.G, 2.37 due to Debian Stable, 2.21 due to oldstable, 2.39 due to Solus holding back for cgroups v2 support in Docker, etc).