Optimized the cluster loop

[micheles]

With less iterations and using less memory. Here are the figures for usa_small:

# before
| calc_3, maxmem=242.1 GB    | time_sec  | memory_mb | counts  |
|----------------------------+-----------+-----------+---------|
| total classical            | 16_854    | 1_336     | 705     |
| cluster loop               | 13_605    | 267.1     | 705     |
| total postclassical        | 1_135     | 48.0391   | 10      |
| reading rates              | 1_133     | 47.9805   | 10      |
| get_poes                   | 750.1     | 0.0       | 103_320 |
| ClassicalCalculator.run    | 485.5     | 803.0     | 1       |

# after
| calc_6, maxmem=305.3 GB    | time_sec  | memory_mb | counts  |
|----------------------------+-----------+-----------+---------|
| total classical            | 7_786     | 1_119     | 705     |
| cluster loop               | 4_293     | 0.0352    | 705     |
| total postclassical        | 1_130     | 0.0       | 10      |
| reading rates              | 1_129     | 0.0       | 10      |
| get_poes                   | 793.5     | 0.0       | 103_320 |
| ClassicalCalculator.run    | 454.8     | 676.3     | 1       |