Hello Nguyen
Yes, the explanation is correct. Routing will cause the exit interface to be Gi0/3 while the first NAT statement will cause the source address to be translated to 192.168.12.100, which causes the ISP2 router to drop the packet, since it has no route to that destination.
The purpose of the extendible keyword here is to allow multiple OUTSIDE to INSIDE NAT translations to take place to the same inside host address. This is useful because users on the Internet can be routed via either ISP to reach that host. However, having the host reach the Internet via two different ISPs is not a problem that the extendible keyword is used to solve. If you want to achieve this, for redundancy for example, there are other ways to do it such as SDWAN, redundant routing, or LAG for example.
I hope this has been helpful!
Laz