fix(java): don't recreate WASM instance on thread interruption

[andreas-karlsson]

Summary

• RecoveringResolver now distinguishes ChicoryInterruptedException from
  other ChicoryException types — interrupts are rethrown without triggering
  background WASM instance recreation, since the instance isn't actually broken.
• PooledResolver.maintenance() no longer catches RuntimeException. Errors
  propagate to callers instead of being silently swallowed. The close() path
  retains its own catch to ensure all slots are cleaned up.
• Scheduler lambdas in OpenFeatureLocalResolveProvider now have error handling
  so that a transient failure doesn't silently kill the state refresh loop or
  the assign log flush timer.

Test plan

• New RecoveringResolverTest verifying that ChicoryException triggers
  recreation but ChicoryInterruptedException does not
• Existing integration tests pass

🤖 Generated with Claude Code