Language server recompiles CQL files from scratch on every keystroke, causing sluggish editor performance

[raleigh-g-thompson]

Every time a user types in a CQL file — or triggers any language feature such as hover, go-to-definition, or View ELM — the language server recompiles the entire library from scratch. This includes reloading all model definitions and dependent libraries every single time, regardless of whether anything has actually changed.

[raleigh-g-thompson]

The underlying issue is how the CqlCompilation class is used. The hover provider is recreating the cache on every call. Change code so that the cache is shared and preserved between measure runs.