Currently the Agent JWT caching system holds all of the JWTs that have been requested, until they expire.

This issue would convert the cache to one of a specific fixed size.

The size should be measured by the count of the JWTs present.

Upon JWT request by a workload, the cache is accessed. If a suitable JWT exists in the cache, it is returned (the short path of handling the Workload API request). When the cache lacks the required JWT, a new JWT is generated by way of communications to the Server, and as part of returning the JWT to the Workload, the new JWT is inserted into the Agent JWT cache.

When inserting into a non-full cache, an empty slot in the cache accepts the JWT.
When inserting into a full cache, the JWT with the least amount of remaining utility is evicted from the cache, and the new entry is inserted.

The cache may or may not evict expired entries outside of a typical fetch / insert operation depending on team decision / ease of implementation.