gracefully handle and warn about zero-gap between events for frequency estimate

[lasch]

More gracefully handle division by zero in src/aiu_trace_analyzer/pipeline/normalize.py and add user feedback:

Changes made:

1. Added a new TraceWarning called "zero_gap_time" to track events with identical timestamps
2. Added check for gap_time > 0 before calculating frequency
3. When gap_time is 0, the code now:
   • Increments the warning counter via self.warnings["zero_gap_time"].update()
   • Falls back to using dur_freq instead of attempting division by zero

Users will now receive feedback about how many events had identical timestamps, helping them identify potential timestamp issues in their trace data. The warning message explains that this causes unreliable interval-based frequency calculation.

[WarningRan]

LGTM.

Fix is correct and well-scoped. gap_time can't be negative (sorted upstream), dur_freq fallback is safe (dur == 0 already filtered by sane_event()). Warning follows existing pattern.