Improved performance for ProtobufHttpMessageConverter

[koo-taejin]

Improved performance for ProtobufHttpMessageConverter

Currently, when a binary message in the form of ProtoBuf is received, data is created using the Builder via ProtobufHttpMessageConverter.
When Protobuf creating individual com.google.protobuf.Message. it creates a PROTO_MESSAGE parseFrom(java.io.InputStream input, com.google.protobuf.ExtensionRegistryLite extensionRegistry) method
If use this method, ProtobufHttpMessageConverter can reduce creating unnecessary Builder and executing merge method process.
(And I have confirmed that grpc-java use this method when creating 'com.google.protobuf.Message'.)

If change the object creation to the parseFrom method, I made a simple test and confirmed that the TPS increased by about 5% in several simple performance tests.
(I have confirmed that the slope of increasing heap improved.)

• as-is

h2load http://localhost:8080/user/proto -d ./proto.txt --header 'Content-Type: application/x-protobuf;charset=UTF-8'  -n 100000 -c 1 -m 1
starting benchmark...
spawning thread #0: 1 total client(s). 100000 total requests
Application protocol: h2c
progress: 10% done
progress: 20% done
progress: 30% done
progress: 40% done
progress: 50% done
progress: 60% done
progress: 70% done
progress: 80% done
progress: 90% done
progress: 100% done

finished in 24.07s, 4155.29 req/s, 146.15KB/s
requests: 100000 total, 100000 started, 100000 done, 100000 succeeded, 0 failed, 0 errored, 0 timeout
status codes: 100000 2xx, 0 3xx, 0 4xx, 0 5xx
traffic: 3.43MB (3601612) total, 488.93KB (500660) headers (space savings 96.26%), 390.63KB (400000) data
                     min         max         mean         sd        +/- sd
time for request:      187us     13.12ms       234us       105us    94.49%
time for connect:      301us       301us       301us         0us   100.00%
time to 1st byte:     2.11ms      2.11ms      2.11ms         0us   100.00%
req/s           :    4155.31     4155.31     4155.31        0.00   100.00%

• to-be

h2load http://localhost:8080/user/proto -d ./proto.txt --header 'Content-Type: application/x-protobuf;charset=UTF-8'  -n 100000 -c 1 -m 1
starting benchmark...
spawning thread #0: 1 total client(s). 100000 total requests
Application protocol: h2c
progress: 10% done
progress: 20% done
progress: 30% done
progress: 40% done
progress: 50% done
progress: 60% done
progress: 70% done
progress: 80% done
progress: 90% done
progress: 100% done

finished in 23.31s, 4289.18 req/s, 150.86KB/s
requests: 100000 total, 100000 started, 100000 done, 100000 succeeded, 0 failed, 0 errored, 0 timeout
status codes: 100000 2xx, 0 3xx, 0 4xx, 0 5xx
traffic: 3.43MB (3601566) total, 488.88KB (500614) headers (space savings 96.26%), 390.63KB (400000) data
                     min         max         mean         sd        +/- sd
time for request:      177us    274.17ms       227us       876us    99.97%
time for connect:      331us       331us       331us         0us   100.00%
time to 1st byte:   254.55ms    254.55ms    254.55ms         0us   100.00%
req/s           :    4289.20     4289.20     4289.20        0.00   100.00%

Please feel free to let me know if I misjudged anything. :)

[snicoll]

@koo-taejin for the record, there's no need to open both an issue and a PR as one supersedes the other.

Closing in favor of PR #29496

[koo-taejin]

@snicoll Thank you for letting me know. :)