perf: tag the three perf-adjacent correctness bugs with the new performance label

[jordanrburger]

Problem

Three existing open issues are functionally performance bugs but are currently only labeled as correctness/bug:

• kb_pre_filter: naive datetime DST drift + rglob follows symlinks (potential infinite loop) #53 — kb_pre_filter rglob follows symlinks and can hit infinite-loop walks. Pure perf/availability impact if a stray symlink loop exists.
• schedule_tick.run() holds fcntl lock during 48s network probe #60 — schedule_tick.run() holds the fcntl lock during a 48 s network probe. Blocks every other tick / fire-now for nearly a minute.
• connector_log: no rotation or size cap (unbounded growth) #66 — connector_log has no rotation or size cap. Direct cause of perf: schedule_tick re-reads entire tracker JSONLs every 5 min #73 getting slower over time.

Without a performance label, these don't show up alongside this audit batch.

Fix

• ✅ Created the performance label (color 1d76db, "CPU/memory/wall-clock optimization").
• Re-tag kb_pre_filter: naive datetime DST drift + rglob follows symlinks (potential infinite loop) #53, schedule_tick.run() holds fcntl lock during 48s network probe #60, connector_log: no rotation or size cap (unbounded growth) #66 with performance so all hot-path work is searchable under one label.

This issue tracks that bookkeeping and can be closed once the labels are applied.