Define table size per scaling test

[jimmyzhai]

This pull request adds a new header file defines.h for constants definitions. It defines different table sizes for outbound routing and ca_to_pa mapping according to hero-implementation-details.

The hardcoded table sizes of outbound routing and ca_to_pa mapping, whose size 4M and 8M respectively (from #592) take up too much memory by default. This memory pressure could make low-RAM KVM DPU crash and be unusable with sonic mgmt test.

The below shows the dash pipeline memory usage of the container simple_switch with different scale number of routing and mapping.

1. 4M/8M 
junhuazhai@junhuazhai-dev-vm:~/workspace/DASH/dash-pipeline$ docker stats --no-stream simple_switch-junhuazhai
CONTAINER ID   NAME                       CPU %     MEM USAGE / LIMIT     MEM %     NET I/O   BLOCK I/O   PIDS
00016dc28662   simple_switch-junhuazhai   0.08%     1.799GiB / 62.78GiB   2.87%     0B / 0B   0B / 0B     20

2. 4K/8K
junhuazhai@junhuazhai-dev-vm:~/workspace/DASH/dash-pipeline$ docker stats --no-stream simple_switch-junhuazhai
CONTAINER ID   NAME                       CPU %     MEM USAGE / LIMIT     MEM %     NET I/O   BLOCK I/O   PIDS
bcb858280193   simple_switch-junhuazhai   0.08%     46.63MiB / 62.78GiB   0.07%     0B / 0B   0B / 0B     20

[r12f]

will be better to use TABLE_BABY_HERO_SCALE

[mgheorghe]

TABLE_BABY_HERO_SCALE i believe it is in the range of 80K to 320K can we at least have the default limits increased to that

[jimmyzhai]

Use TABLE_HERO_SCALE/TABLE_BABY_HERO_SCALE for hero/baby-hero respectively.

[r12f]

thanks a lot Junhua for making this change!

[r12f]

LGTM, thanks for getting the numbers and macro align with the spec!

[r12f]

@mgheorghe , would you like to take another look?