[AI] Track embedding semantic extraction model usage in the `embed-document` worker

[rmolinamir]

MDocument uses gpt-4o-mini to chunk the document, but we're not tracking the usage. We need to fix this or remove metadata extraction.

See:

journl/apps/web/src/app/api/supabase/embed-document/route.ts

Line 117 in 656ead7

model_id: model.modelId,

[rmolinamir]

See:

journl/apps/web/src/app/api/supabase/embed-document/route.ts

Line 16 in 6428018

// ! TODO: Mastra's MDocument uses `gpt-4o-mini` to chunk the document, but we're not tracking the usage.

Disabled for now.