Agent execution tracing

Capturing an autonomous agent's tool calls, decisions, and pipeline steps as inspectable, queryable spans so its activity can be navigated and audited like a graph.

grounded in: gen_activity.py ('sanitized execution-activity feed (router/autopost/self-improve/deploys)', commit 0ffb3b9) plus the trend theme of turning 'agent activity into queryable or spatially-explorable grap