v0.7.0 release: geo-tagging and brain UI fixes

Launched geo-tagging across transcripts, episodes, and the scheduler. Location data is now captured via GPS on write, aggregated per episode window, and surfaced through retrieval paths. A new PlaceAbility handles named places with a cosine-supersede contradiction policy in the knowledge graph.

Extended the subconscious worker to step seven with GeoPatternProcessor for single-pass LLM location-based pattern extraction. Infrastructure replaced raw Nominatim calls with geopy and introduced geo_utils for distance and speed estimation. Architecture documentation was updated to reflect the expanded DISCOVERABLE list, glossary additions, and departure-time reminder logic.

Restored three Brain UI regressions from the recent redesign: the providers-only lock now correctly blocks navigation until a provider is configured, model lists auto-load on modal open and platform switches, and the sidebar logo properly links back to chat. Updated web-interface docs to match the actual providers-only enforcement behavior.

Tagged v0.7.0 release and removed a duplicate docstring line in subconscious_worker.py.

• Transcript and episode tables gain location columns with auto-GPS tagging on write

• GeoPatternProcessor added as step seven of the subconscious worker for behavioral pattern extraction

• Providers-only lock restored with sidebar locks, command palette disabling, and auto-unlock on first provider save

• New PlaceAbility supports named place management with cosine-supersede contradiction policy

• Departure reminders use layered speed estimation: historical median, heartbeat deltas, or 30 km/h default