Hashmap improvements

[t-b]

• The rehashing is still done single-threaded. We could multithread that as we know that a single source line of the small hashmap is only mapped to two in the large hashmap (due to 1 bit more of the DJB hash we use).
• Extend for INT64/UINT64 without precision loss for value
• Rehashing currently always doubles the size if required. But if we have a load factor >> 1 this can still be not enough. So let's calculate the required new size 2^x as something like x >= log_2(numEntries / MAX_LOAD_FACTOR) instead of just using x + 1.