[Java feature server] Converge ServingService API to make Python and Java feature servers consistent

[pyalex]

What this PR does / why we need it:

Currently, Java Feature Server and Python Feature Server provide two different client APIs. This PR aimed to fix that.

1. Changes to proto API: use GetOnlineFeatureRequest (that was created for and currently used by Python FS) as input in GetOnlineFeatures method. That also means that Java FS will support requests by feature service.
2. GetOnlineFeatureResponseV2 is proposed as optimized version of GetOnlineFeatureResponse and was designed symmetrically to GetOnlineFeatureRequest. According to my evaluation serialization/deserialization of the new message is 3x times faster (on 100 rows). That difference increases with a growing number of rows and features. New message also optimized for compaction (message size), which will lead to faster network transmission.

What was updated:

1. Java Feature Server
2. Java SDK
3. Go SDK

What's not covered in this PR:

1. Python FS to return new response

Some clean up:

• "Feature Table" is replaced to "Feature View"
• Suffix "V" is dropped when there are no other versions

Which issue(s) this PR fixes:

Fixes #

Does this PR introduce a user-facing change?:

Java feature server supports retrieving features by feature service name.

ServingService proto API is refactored:
GetOnlineFeaturesResponse now has new argument `request_context` for passing input to ODFVs (previously this inputs were mixed into `entities`
GetOnlineFeaturesResponseV2 is refactored version of response optimized for fast serialization / deserialization and high compaction

[codecov-commenter]

Codecov Report

Merging #2166 (50fc6cb) into master (e435d92) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #2166      +/-   ##
==========================================
+ Coverage   84.57%   84.58%   +0.01%     
==========================================
  Files         102      102              
  Lines        8195     8201       +6     
==========================================
+ Hits         6931     6937       +6     
  Misses       1264     1264              

Flag	Coverage Δ
integrationtests	74.27% <100.00%> (-0.27%)	⬇️
unittests	58.99% <40.00%> (-0.05%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
sdk/python/feast/feature_store.py	91.39% <100.00%> (ø)
sdk/python/feast/infra/online_stores/datastore.py	80.79% <0.00%> (+0.67%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e435d92...50fc6cb. Read the comment docs.

[pyalex]

Benchmarks were also added in Java integration tests
Results:

Test 100 rows; Full FS took (min): 32 ms
Test 100 rows; Full FS took (avg): 40.117172 ms
Test 100 rows; Full FS took (median): 36.000000 ms
Test 100 rows; Full FS took (95p): 64.000000 ms
Test 100 rows; Full FS took (99p): 82.000000 ms

Test 100 rows; 50 features took (min): 17 ms
Test 100 rows; 50 features took (avg): 20.864646 ms
Test 100 rows; 50 features took (median): 19.000000 ms
Test 100 rows; 50 features took (95p): 28.000000 ms
Test 100 rows; 50 features took (99p): 55.000000 ms

Test 100 rows; 10 features took (min): 3 ms
Test 100 rows; 10 features took (avg): 5.576768 ms
Test 100 rows; 10 features took (median): 5.000000 ms
Test 100 rows; 10 features took (95p): 15.000000 ms
Test 100 rows; 10 features took (99p): 18.000000 ms

Test 100 rows; 100 features took (min): 17 ms
Test 100 rows; 100 features took (avg): 20.643434 ms
Test 100 rows; 100 features took (median): 19.000000 ms
Test 100 rows; 100 features took (95p): 26.000000 ms
Test 100 rows; 100 features took (99p): 55.000000 ms

Full FS == feature service with 250 features

[adchia]

also change the release notes to say that the java FS supports feature services now?

[adchia]

to clarify, this is the "request_data" that we had before right?

[pyalex]

To my knowledge, there was no "request_data" in protos or Python API. Not sure what you're referring to

[adchia]

This was used for the on demand feature views. You had touched some of this in the transformation logic (it used to be passed in as part of the entity row). I think this is the same.

Can you also add a new issue to migrate the python logic for this too?

[adchia]

/lgtm

[feast-ci-bot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: adchia, pyalex

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [adchia]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment