Cache isolated analysis results to disk in order to speed up session refreshes

[JonatanWaern]

At the moment, there is a significant 'hump' when starting up the language server targeting some device file, where most of it consists the dls re-parsing a large number of commoncode files which rarely change. We could store isolated analysises to disk (together with a file hash) and try to re-use these when an analysis is missing.

[JonatanWaern]

Improvements to path-resolver caching has made this less relevant