Investigate CC performance hotspots

[bamboo]

• What are the main hotspots for Parallel Store (cache miss)?
• What are the main hotspots for Parallel Load (cache hit)?
• How much data duplication is still left? See also internal discussion
• Can we gain more performance by deduplicating other object types beside strings? e.g. Configuration, FileCollection, etc
• Is the strategy we're using for IO, i.e. writing project files in parallel, good enough?
• Can a fast compression scheme like zstd improve performance?