EVM/Node: GC keeps number of chunks

The garbage collector doesn't do what it was supposed to. The history_to_keep_in_seconds is rather an gc_frequency_in_seconds, e.g. it triggers the gc every N days instead of keeping N days. The merge request fixes this problem by using a different (simpler) approach.

The node has now 2 gc parameters:

1. Split frequency in seconds
2. Number of chunks

If you want to keep 7 days of history you'd do:

{
  "split_frequency_in_seconds": 86_400,
  "number_of_chunks": 7
}