GUACAMOLE-2221: fix RDP busy-loop on transport failure

[pwgcz]

When an RDP connection hits a transport-level
failure (for example, a certificate error), the
inner event loop in rdp.c can spin at 100% CPU.

This happens because the error state in
wait_result gets overwritten. When the issue
occurs, guac_rdp_handle_events() sets wait_result
to -1, but the loop doesn’t exit immediately. In
the next iteration, the while condition assigns
wait_result from rdp_guac_client_wait_for_events()
effectively clearing the error flag. As a result,
the loop never observes the transport failure and
continues polling, spinning the CPU.

Fix: break out of the loop immediately after
guac_rdp_handle_events() fails, so we stop
processing when FreeRDP reports a connection
error.

[necouchman]

@pwgcz As discussed in the Jira ticket, I don't think we want to simply revert the previous patch, as that just reintroduces the issue that the commit resolves. You had suggested an alternate approach in the Jira issue - you're welcome to update this PR with that approach and we can see if that addresses the issue without (re)introducing other issues.