A trunk group is local to the router, at least if you by "trunk group" is referring to the configuration element in IOS. If you mean a trunk group as in your service provider has configured your PRI's in a group that has no relevance to what router you connect the physical ISDN line to.
To do this you would create two SIP trunks in CM, or whatever control protocol that you use. It does not have to be SIP, but is for sure the recommended option to use. Then you put these two trunks into a route group and set the Distribution Algorithm to either Circular or Top Down depending on what you want. Put this route group into an route list and then point your route patterns to this route list.
For both the SIP trunks and route list make sure that your set the check box for Run on all nodes. This is to get the best possible traffic pattern and resiliency.
In the gateway it is recommended to use a server group on the outbound dial peer to CM so that you don't need to maintain multiple dial peers. For detailed information on how to setup call routing in IOS have a look at this excellent document In Depth Explanation of Cisco IOS and IOS-XE Call Routing.
The recommended option is to use Sip, MGCP work well for such scenarios but its these days we dont recommend MGCP much.
When using SIP protocol for the gateway,
connect the line 1 to Gateway A, connect the line 2 to gateway B. On CUCM add trunks to Gateway A and B. you can add both gateways on are Route Group and use algorithm circular , an example mentioned below.
In both case if one PRI line goes down , to send all the calls through the second gateway you need to use a work around which is
option 1:- on service parameter "Stop Routing on Unallocated Number Flag Required Field " make it True. this is not actually a solution but it is a work as making it true when you dial wrong numbers this gets applied.
option 2:- have a look on https://community.cisco.com/t5/ip-telephony-and-phones/cucm-gt-pri-gateway-over-sip-trunk-failover-when-pri-goes-down/td-p/2645889
MGCP
The failover will work well on MGCP as CUCM control the PRI Lines. So when one line goes down CUCM send all the calls through second PRI without any configurations changes if you have RG or RL configured as mentioned above.
Typically you would not use any overlay functionality for Active/Standby. The typical setup would be to configure each router as an independent gateway and let the functionality in CM take care of call distribution for outbound calls.
On your other question, inbound calls are handled by your service provider. It’s up to them to define what happens when a circuit is down. Normally they would configure the two circuits in a group and distribute calls amongst them. This would handle call distribution in the event one gateway or circuit is down.
