perf(db): cache ProcessUID::own in memory

[DaniPopes]

System::new_with_specifics is fairly expensive since it will read and parse all of /proc, on my machine it takes ~350ms.

I think caching our own PID is valid since we only use start_time.

After opening this PR I realized we can filter the parsing step to only parse /proc/<id>, and this takes it down to a few microseconds (I can't even see it with samply anymore); I think caching is still worth it just to not deal with any files after parsing it once

(This is called at least twice, in DatabaseEnv::open and StaticFileProvider::new)

[mattsse]

I think caching our own PID is valid since we only use start_time.

I believe this is correct.

nice