Cache cluster schema (knowledge?) to prevent multiple discovery queries on every new session

[danield137]

If allowed (RTI_MCP_KUSTO_CLUSTER_CACHE,RTI_MCP_KUSTO_CLUSTER_CACHE_PATH), whenever a schema call is made, cache the results in a cache folder (RTI_MCP_KUSTO_CLUSTER_CACHE_PATH or, by default, user app data), with a timestamp.
Next time a similar request is made, return from cache (let's make the staleness time about ~1d by default, and allow users to configure their own staleness time, per cluster/database pair (for reference, see how known_servoces is structred).
add tests.