perf: Have read relationships use dynamic index forcing

[josephschorr]

This should ensure that, for recognized query shapes, the best index and sort order is selected for ReadRels

Testing

Unit, datatstore and steelthread tests

[codecov]

Codecov Report

❌ Patch coverage is 60.46512% with 17 lines in your changes missing coverage. Please review.
✅ Project coverage is 59.15%. Comparing base (9029960) to head (45bdbf6).

Files with missing lines	Patch %	Lines
internal/datastore/common/sql.go	37.50%	3 Missing and 2 partials ⚠️
internal/datastore/crdb/reader.go	33.34%	2 Missing and 2 partials ⚠️
internal/datastore/crdb/schema/forcedindex.go	0.00%	2 Missing ⚠️
internal/datastore/crdb/schema/indexes.go	88.89%	1 Missing and 1 partial ⚠️
internal/datastore/postgres/schema/indexes.go	0.00%	2 Missing ⚠️
internal/datastore/spanner/indexes.go	0.00%	2 Missing ⚠️

❌ Your project check has failed because the head coverage (59.15%) is below the target coverage (75.00%). You can increase the head coverage or adjust the target coverage.

❗ There is a different number of reports uploaded between BASE (9029960) and HEAD (45bdbf6). Click for more details.

HEAD has 21 uploads less than BASE

Flag	BASE (9029960)	HEAD (45bdbf6)
	47	26

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #2632       +/-   ##
===========================================
- Coverage   79.52%   59.15%   -20.36%     
===========================================
  Files         454      412       -42     
  Lines       47138    43740     -3398     
===========================================
- Hits        37481    25870    -11611     
- Misses       6894    15369     +8475     
+ Partials     2763     2501      -262     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[miparnisari]

Shall we rename this PR from feat: ... to perf: ... since it's not a new feature but rather a performance improvement?

[miparnisari]

why give the caller a new option, when unsorted: lets the underlying datastore choose the order exists? or should we update unsorted to just say no order at all?

[josephschorr]

Because sometimes the caller knows better and letting it choose is important in those instances

[miparnisari]

By reading https://buf.build/authzed/api/docs/main:authzed.api.v1#authzed.api.v1.PermissionsService.ReadRelationships it doesn't seem that we ever promised a particular return order. I'm curious why this code was the way it is with options.ByResource? was this an undocumented feature or just an accident?

can we update the API to be explicit that callers shouldn't expect any order?

[josephschorr]

Its necessary for cursoring; we need some stable order in order to cursor

[miparnisari]

ok; should we update the API docs to say something like "the results will be sorted but the order will depend on the filter passed"?