rework prune rewrite with iterators

[antiochp]

This PR refactors the handling of pos_to_rm when pruning during compaction.
We used to pass a slice &[u64] around and this has been reworked to accept an IntoIterator<Item = u64>.
This allows us to avoid collect() in various places with intermediate memory allocation.

Split save_prune() into two functions for flexibility so we can pass an iterator into the earlier write_tmp_pruned() without needing &mut self (as the iterator passed in has a reference to immutable self).

[antiochp]

This is purely a refactor. We have decent test coverage. Testing locally to ensure we can still prune/compact successfully.
Will merge if local testing is good.

[antiochp]

20210219 11:37:56.842 DEBUG grin_chain::txhashset::txhashset - txhashset: starting compaction...
20210219 11:38:23.665 DEBUG grin_chain::txhashset::txhashset - txhashset: check_compact output mmr backend...
20210219 11:38:26.931 DEBUG grin_chain::txhashset::txhashset - txhashset: check_compact rangeproof mmr backend...
20210219 11:38:31.452 DEBUG grin_chain::txhashset::txhashset - txhashset: ... compaction finished