Revisit files and compactions metrics in MerkleDb

This is a follow-up for https://github.com/hiero-ledger/hiero-consensus-node/issues/23987 and related issues.

MerkleDb metrics are somewhat messy today. There are several categories (files, reads, compactions, etc.), some labels are encoded in metric names (we now have support for explicit metric labels). Besides that, with data file compaction v3, some metrics will become obsolete or questionable, like time spent compacting some level - there may be many compaction jobs handling the same level, which one of them to report?

"Files" metrics need a rework, too. File levels should be extracted to metric labels, for example.