logging: rtt: Improve synchronous mode

RTT backend was treating synchronous mode (LOG_IMMEDIATE) and panic
mode in the same way. That lead to decreased bandwidth since after
each transfer operation backend was pending until RTT data is read by
the host. It is vital only in panic mode to ensure that device do not
reset before all data is read by the host. In synchronous mode that
degrades performance significantly.

Added distinction between those two modes.

Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>