container.list_objects with prefix and the local driver is very slow when a lot of files are stored

## Summary

I have a large storage that I'm using libcloud to manage - in this specific case it's around 500,000 objects, and I've noticed that listing object using the local driver and a prefix is very slow (took almost 2 minutes to return 90 objects).

## Detailed Information
In the local driver implementation when calling list_objects with a prefix - we firstly get all the objects and serialize them and just after that we are filtering by the prefix.

Found on python3.6.8 with libcloud 3.2.0 and ubuntu 16:04

The issue points to the local driver's container.list_objects path; start there and inspect how objects are serialized before prefix filtering. Reproduce the report with a large local object set and a prefix, then verify that matching objects are returned without the reported multi-minute delay.

Vom Indexierungsmodell aus dem Issue-Text verfasst.