Ensure EVM database integrity and consistency with DVM database

[sieniven]

Summary

This PR fixes issues that cause database integrity issues on the EVM side, resulting in a corrupted rocksdb.

1. ADD EVM db flush in DVM's FlushStateToDisk pipeline to keep EVM and DVM DB flushes as consistent as possible.
2. Expose EVM db flush on FFI.
3. Propagate errors on EVM services shutdown pipeline correctly and do not panic the node on failed DB flush.
4. Add pipeline in init to check for DVM and EVM database state inconsistencies. If state inconsistency is present, the startup pipeline will trigger re-indexing pipeline.
5. Propagate errors on RPC servers shutdown pipeline correctly instead of panicking the node. Previous pipeline panics the node early which prevents rocksdb from properly flushing in-memory buffer on shutdown, causing db corruption on subsequent startups.
6. Propagate errors from RPC servers correctly and set initialization failure instead of panicking the node on start up.

Implications

• Storage

  • Database reindex required
  • Database reindex optional
  • Database reindex not required
  • None

• Consensus

  • Network upgrade required
  • Includes backward compatible changes
  • Includes consensus workarounds
  • Includes consensus refactors
  • None

[sieniven]

EVM db flush is shifted out of every tip change on the rust side with commit c86861f. Instead, the EVM db flush is now exposed on the FFI, and is called in the same pipeline on the DVM side to keep the DVM and EVM databases consistent in FlushStateToDisk.