Use our "rich error" facility to propagate error reported by the stub to
the user. lldb-server reports rich launch errors as of D133352.
To make this easier to implement, and reduce code duplication, I have
moved the vRun/A/qLaunchSuccess handling into a single
GDBRemoteCommunicationClient function.