Allow applications to selectively ignore cache limit with in-memory configuration

The in-memory storage engine currently returns WT_CACHE_FULL if an application attempts to do any insert or update operation when the cache is at capacity.

There are use cases where applications might want to allow specific operations to complete even though the cache is full. The particular use case in mind is an application that tracks metadata in a WiredTiger table and requires that inserts/updates to the metadata are allowed for correctness.