Custom metric API example

[rmccorm4]

Adds an example of adding a custom metric family global to the backend with BackendState, and per-model metrics added to the global metric family by each ModelState via labels:

root@RMCCORMICK:/mnt/jira/3557-metrics-example/server/qa/L0_backend_identity# grep byte_size client.log
# HELP input_byte_size_counter Cumulative input_byte_size across all identity model requests
# TYPE input_byte_size_counter counter
input_byte_size_counter{model="identity_bytes",version="1"} 0.000000
input_byte_size_counter{model="identity_fp32",version="1"} 0.000000
input_byte_size_counter{model="identity_nobatch_int8",version="1"} 0.000000
input_byte_size_counter{model="identity_uint32",version="1"} 64.000000

Updated unit test and documentation here: triton-inference-server/server#4211