Bump actions, fix cache, move to chainguard, simplify docker flow

[clux]

• Upgrading actions like in Try upgrading all the actions and switch to chainguard version-rs#14
• Tried similar cache attempt except, but it's still not caching volume layers so moved back to manual building
• Split docker job into two (otel + base) and using metadata action instead of our the old prep action which is nicer
• Changing dockerfile setup to builder into chainguard (and moving out labels, populated by metadata action)
• Minor cleanups here and there

[clux]

now everything builds in 1m again.
kind of sad that the gha docker cache still does not cache mounted cache data, otherwise we could use the normal builder pattern, but i'm ok with invoking the docker build manually (or letting the user run cargo build --target directly if possible) if it means actually having caching work.
anyway this was a bit of experimentation for kube-rs/website#14