EntryWrappingInterceptor should not wrap entries on the originator in non-tx mode

EntryWrappingInterceptor assumes that lock delegation is only enabled in dist caches, so in a repl cache it will wrap the entry twice. This may mean that the entry is also written to the data container twice.

I have tried changing this assumption in the context of ISPN-3289, but it caused some failures in ClusteredCacheWithRamDirIndexManagerTest and ClusteredCacheWithInfinispanDirectoryTest. The query interceptor may be assuming that the entries are committed to the data container when ctx.isOriginLocal() == true, so it warrants a more thorough investigation.