Speeding up count_ruptures in event_based

[micheles]

Part of #9066. This is tripling the speed of an example calculation I have:

# before
| calc_15024, maxmem=5.5 GB  | time_sec | memory_mb | counts |
|----------------------------+----------+-----------+--------|
| total sample_ruptures      | 196.3    | 30.4      | 42     |
| EventBasedCalculator.run   | 368.8    | 603.4     | 1      |

# after
| calc_15023, maxmem=6.1 GB  | time_sec | memory_mb | counts |
|----------------------------+----------+-----------+--------|
| total sample_ruptures      | 198.1    | 6.00000   | 42     |
| EventBasedCalculator.run   | 137.1    | 603.2     | 1      |

Here is the speedup in .iter_ruptures being called 3x less times:

| ncalls    | cumtime | path                                          |
|-----------+---------+-----------------------------------------------|
| 2752431   | 521.475 | kite_fault.py:125(iter_ruptures)              |
| 917477    | 176.358 | kite_fault.py:163(iter_ruptures)              |

In @CatalinaYepes calculation for Domenican Republic (#54890) the speedup in counting the ruptures is 4x,
from 119m to 29.5m.