Add discovery mode context injection and aggregate_speculative_facets()
When no facets are defined, the facet classifier now receives explicit discovery-mode
context instructing it to name observed contexts with specific, descriptive labels.
Previously it ran with zero facet context, producing unconstrained speculative output.
Also adds aggregate_speculative_facets() to think/facets.py, which scans per-segment
agents/facets.json files across days, counts facet name frequency, and returns the
top 8 by count with sample activities. Used by onboarding agents to suggest journal
organization to the user.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>