Optimize semantic validation

[jsoriano]

Semantic validators read and parse the same files multiple times. For example we have ~9 semantic validators that parse fields files, what in some cases takes most of the validation time.

I discussed about this problem to copilot/claude, and we decided on some refactors, that basically Claude implemented.

The changes are basically adding caches at the file system level, and when unmarshalling and getting values from files.

The average improvement when tested with the packages in the integrations repository is around 22%, with 88% reduction in unmarshall operations (as they are cached), and around 16% reduction in total allocations and GC activity.

Cache entries never expire, but the cache itself only lives during the validation, and then is cleaned up by the GC.

[elasticmachine]

💚 Build Succeeded

• Buildkite Build
• Commit: 5a4c1e7

cc @jsoriano