Plasma should not loop indefinitely when writing to disk fails continually

When a golang memory allocation fails continuously, plasma keeps retrying to flush data to disk and fails with error .

Unable to write - err write

Plasma should fail after a set number of retries with a long timeout and possibly assert fail to restart the service and get the golang memory issue to be resolved on restart.

Additionally, when there is a failure, along with error, consider logging more information like:

1. The size of write requested
2. Details of memory in various allocation slabs similar to what is logged in couchbase.log when an OOM kill is observed