Cache Broken Connections

[ddimaria]

Basic description/background

When a connection to a datasource in the connection service is not possible, it can time out in some circumstances (e.g. ssh connections that don't respond). While this is expected in some cases, we can optimize the connection service to explicitly avoid retrying for a period of time (TBD).

For schema API calls, add a url parameter ?cache_error=true to invoke this system.

In the connection service, convert the cache to an enum instead of a struct to capture the different states:

Current:

struct Cache {
    schema: Arc<Mutex<MemoryCache<Uuid, Schema>>>,
}

Proposed:

enum SchemaCache {
    Success(Schema),
    Error(String),
}

struct Cache {
    schema: Arc<Mutex<MemoryCache<Uuid, SchemaCache>>>,
}