Open a new file based db.

Use full disk kv store to remove all duplicates - it should have low heap memory footprint but lots of I/O interactions.

MemoryFilter uses bitset in-memory filters to remove duplicates.

MemoryLRU keeps in memory the last x items and filter with a look-back probabilistic window with fixed size.

Uses go standard map without eviction - grows linearly with the items number.

50Mb.

FileDB - represents a file db implementation.