feat: add feature flag endpoint

[daniel-codecov]

Purpose/Motivation

What is the feature? Why is this being done?

Dependent on codecov/shared#205.

Adds an endpoint at /internal/features which will accept a request body of

{
    "feature_flags": [
        "my_feature_1",
        "my_feature_2",
        "my_feature_3,
        "my_feature_4"
    ],
    "identifier_data": {
        "email": "daniel.yu@sentry.io",
        "user_id": 123,
        "org_id": 1,
        "repo_id": 123213
    }
}

and spit out a response with all the flag evaluations for that given user based on their identifier_data:

{
    "my_feature_1": 4,
    "my_feature_2": false,
    "my_feature_3": true,
    "my_feature_4": false
}

Mostly copies over the core logic for https://github.com/codecov/shared/blob/f7768f29cb0084947d4eb0c42db6c747ad3ee3f0/shared/rollouts/__init__.py#L27 except with its own caching strategy using Redis. The caching strategy here:

1. Attempt to fetch all flags being requested from redis cache
2. Of those flags with a cache-hit, evaluate them (figure out what variant they get)
3. Of the flags with a cache-miss, bulk fetch them all from the db. Evaluate all these flags, and also store them in Redis which has 5 min TTL.
4. Its actually necessary to not refresh the TTL of cache-hit keys since if they did refresh, the flags values would be forever stale if the flag is updated in django admin

Also adds tests for the endpoint

Legal Boilerplate

Look, I get it. The entity doing business as "Sentry" was incorporated in the State of Delaware in 2015 as Functional Software, Inc. In 2022 this entity acquired Codecov and as result Sentry is going to need some rights from me in order to utilize my contributions in this PR. So here's the deal: I retain all rights, title and interest in and to my contributions, and by keeping this boilerplate intact I confirm that Sentry can use, modify, copy, and redistribute my contributions, under Sentry's choice of terms.

[codecov]

Codecov Report

Attention: Patch coverage is 95.29412% with 4 lines in your changes are missing coverage. Please review.

Project coverage is 95.79%. Comparing base (8601246) to head (f9f3274).

✅ All tests successful. No failed tests found.

Files	Patch %	Lines
api/internal/feature/views.py	96.49%	2 Missing ⚠️
api/internal/feature/helpers.py	92.85%	1 Missing ⚠️
codecov/admin.py	75.00%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##            main    #531     +/-   ##
=======================================
- Coverage   95.80   95.79   -0.01     
=======================================
  Files        774     777      +3     
  Lines      17124   17208     +84     
=======================================
+ Hits       16404   16484     +80     
- Misses       720     724      +4     

Flag	Coverage Δ
unit	91.51% <95.29%> (+0.01%)	⬆️
unit-latest-uploader	91.51% <95.29%> (+0.01%)	⬆️

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

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[matt-codecov]

rather than storing the FeatureFlag in a redis cache, could you go a layer higher and store the list of flag results in the cache based on the input ID info? or we could skip redis for now - i'm not sure whether going to redis on every request will be significantly faster than going to the db on every request

in any case, i think we should try to avoid the duplicated implementation of the feature bucketing logic if we can. is there a reason why we can't create a Feature() and call .check_value()? its caching may not be useful, but i don't think it would be harmful

[daniel-codecov]

rather than storing the FeatureFlag in a redis cache, could you go a layer higher and store the list of flag results in the cache based on the input ID info?

I feel like if we cache based on the user info, we would basically never get a cache hit since gazebo should only hit the endpoint at some interval of like 5 min

[daniel-codecov]

in any case, i think we should try to avoid the duplicated implementation of the feature bucketing logic if we can. is there a reason why we can't create a Feature() and call .check_value()? its caching may not be useful, but i don't think it would be harmful

I can make it work, just need to do a minor rework of the class

[codecov-qa]

Codecov Report

Attention: Patch coverage is 95.29412% with 4 lines in your changes are missing coverage. Please review.

Project coverage is 91.51%. Comparing base (8601246) to head (f9f3274).

✅ All tests successful. No failed tests found.

Files	Patch %	Lines
api/internal/feature/views.py	96.49%	2 Missing ⚠️
api/internal/feature/helpers.py	92.85%	1 Missing ⚠️
codecov/admin.py	75.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #531      +/-   ##
==========================================
+ Coverage   91.49%   91.51%   +0.01%     
==========================================
  Files         599      602       +3     
  Lines       16268    16352      +84     
==========================================
+ Hits        14885    14965      +80     
- Misses       1383     1387       +4     

Flag	Coverage Δ
unit	91.51% <95.29%> (+0.01%)	⬆️
unit-latest-uploader	91.51% <95.29%> (+0.01%)	⬆️

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

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[codecov-public-qa]

Codecov Report

Attention: Patch coverage is 95.29412% with 4 lines in your changes are missing coverage. Please review.

Project coverage is 91.51%. Comparing base (8601246) to head (f9f3274).

✅ All tests successful. No failed tests found ☺️

@@            Coverage Diff             @@
##             main     #531      +/-   ##
==========================================
+ Coverage   91.49%   91.51%   +0.01%     
==========================================
  Files         599      602       +3     
  Lines       16268    16352      +84     
==========================================
+ Hits        14885    14965      +80     
- Misses       1383     1387       +4     

Flag	Coverage Δ
unit	91.51% <95.29%> (+0.01%)	⬆️
unit-latest-uploader	91.51% <95.29%> (+0.01%)	⬆️

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

Files	Coverage Δ
api/internal/feature/serializers.py	100.00% <100.00%> (ø)
api/internal/urls.py	97.29% <100.00%> (+0.07%)	⬆️
api/internal/feature/helpers.py	92.85% <92.85%> (ø)
codecov/admin.py	91.11% <75.00%> (-1.75%)	⬇️
api/internal/feature/views.py	96.49% <96.49%> (ø)