improve robustness to orientdb shutdown after stopping the store

The attached log shows a graceful shutdown on Windows and a message about a corrupt OrientDB because the database.ocf file was zero bytes.

There's a potential flaw in the NPM shutdown logic where an exception while stopping the store means we don't go on and cleanly shutdown OrientDB: https://github.com/sonatype/nexus-oss/blob/nexus-2.11.x/plugins/npm/nexus-npm-repository-plugin/src/main/java/com/bolyuba/nexus/plugin/npm/service/internal/orient/OrientMetadataStoreLifecycle.java#L56

This doesn't explain why the state transition failed ( in the attached log ), but adding a try...finally block here should avoid leaving OrientDB in an odd state when it does happen.