#3265 step 3: function-level PageRank, fan-in and fan-out as new columns #3330

Part of #3265 (features 1 and 3). Depends on #3328.

Build a function-level graph from the resolver's confident fcall edges, using the native gitgalaxy/core/graph_engine.py (no networkx), and write new function_data columns, e.g. func_pagerank, func_fan_in, func_fan_out. No existing file-level column or risk score changes -- the non-breaking option #3237 already recommends.

Decide: whether ambiguous edges count, and with what weight. Whether recursion and call multiplicity feed edge weights (follows from #3327).

Downstream-impact query (#3265 feature 3): expose a traversal ("which functions are reachable upstream of X") in the recorder DB or as a tool query, reading the same edges.

Performance: big repos have hundreds of thousands of functions. Measure with the perf-ab skill on a large real repo before this is on by default.

Done when: the columns are populated across the corpus, the before/after scan time is recorded, and existing columns are unchanged.