Dremio odbc interpretation of flight auth

[jmao-denver]

Three bugs found while working with alternative Flight clients:

• Some clients refuse to use authentication types besides "Bearer" and "Basic" - we should overload "Bearer" to support a nested type so we can instruct clients to specify the type and token in the same call.
• Our handling of Bearer requires that the token be a UUID by mistake - we intended to handle failure to read the UUID as "this isn't an existing session", but instead we fail the entire auth process
• Some clients connect to Handshake as if it wasn't streaming or unary, but send no messages before half-closing. In those cases we assume that the auth process didn't work, but instead we can check to see if an auth header was set instead.