Improve pruning: performance, config and progress logging

[magicxyyz]

part of NIT-2541

This PR:

• reintroduces performance improvement previously removed while fixing upstream geth merge

// note: we are passing data.Root as stateRoot here, to skip the check for stateRoot existence in trie.newTrieReader,
// we already check that when opening state trie and reading the account node
trieID := trie.StorageTrieID(data.Root, key, data.Root)
storageTr, err := trie.NewStateTrie(trieID, sdb.TrieDB())

• enables cleans cache in triedb used for traversing storage tries in dumpRawTrieDescendants - when opening storage trie with trie.NewStateTrie (using the trick above), a node with hash equal data.Root is read twice from triedb (once in when getting a trie.Reader, then when reading the node). Caching the node should save some of KeyValueStore reads.

• adds pruner.Config.Threads option

• adds extra progress logging when a thread traversing storage trie for given account takes longer then 5 minutes

[PlasmaPower]

LGTM