GRE loopback-avoidance mode & GRE-only tunnel services scaling

By default, when a packet is encapsulated on a generic routing encapsulation (GRE) interface, it is sent through the loopback, where a second lookup is performed on the outer IP-GRE header to determine its forwarding path. With loopback avoidance, the packet bypasses this loopback stage and is forwarded immediately after GRE encapsulation in a single pass, preserving forwarding bandwidth and improving efficiency. Loopback avoidance on GRE logical interfaces is supported for the inet, inet6, and bridge families.

Note that GRE logical interface (IFL) scale was limited to 4K per Physical Interface Card (PIC). In 26.2R1, we've increased this scale by dedicating tunnel resources on a PIC exclusively to GRE IFLs, allowing up to 32K GRE IFLs per chassis. Within that overall limit, a maximum of 16K IFLs is supported per interface device (IFD), whether it is a GRE IFD or a physical IFD.