* search: create a separate http endpoint
* search: extract api uri
* search: rename uri
* search: replicate the readiness check
* search: replicate the readiness check metric
* search: update mock
* Search: use SQL search as a fallback when bluge indexing is ongoing
* Search: lint
* Search: feedback fixes - return an empty frame with a special name
* Search: revert readiness check query type
* Search: remove println
* remove sleep, get coffee
* Previews: datasource permissions
* lint
* simplify - force non-null `ds_uids`
* add `canBeDisabled` to search service
* add `IncludeThumbnailsWithEmptyDsUids`
* remove force refresh migration
* refactor main preview service
* add safeguard
* revert ticker interval
* update testdata
* fix test
* add mock search service
* add datasources lookup test
* update migration
* extract ds lookup to its own package to avoid cyclic imports
* lint
* fix dashbaord extract, use the real datasource lookup in tests. IS IT BULLETPROOF YET?!
* fix dashbaord extract, use the real datasource lookup in tests. IS IT BULLETPROOF YET?!
* remove stale log
* consistent casing
* pass context to `createServiceAccount`
* filter out the special grafana ds
