Refactor status collector architecture with structured errors

[dmartinol]

Fixes #2004

• Split the monolithic Collector interface with SyncStatusCollector and APIStatusCollector to pass only the needed one to the processing methods
• Use an Error type to transport condition and phase information to the root, instead of a generic error
• This way, the low-level methods do not need the collectors because they raise the error that carries the desired information

This change is needed to solve #1749 for error paths: some more changes may needed to fix the e2e tests.

[codecov]

Codecov Report

❌ Patch coverage is 28.32765% with 210 lines in your changes missing coverage. Please review.
✅ Project coverage is 48.09%. Comparing base (aa8813f) to head (26c3503).

Files with missing lines	Patch %	Lines
...ator/pkg/mcpregistrystatus/mocks/mock_collector.go	0.00%	77 Missing ⚠️
...thv-operator/controllers/mcpregistry_controller.go	0.00%	65 Missing ⚠️
cmd/thv-operator/pkg/sync/manager.go	33.33%	45 Missing and 1 partial ⚠️
cmd/thv-operator/pkg/registryapi/manager.go	0.00%	17 Missing ⚠️
...thv-operator/pkg/registryapi/mocks/mock_manager.go	0.00%	5 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2016      +/-   ##
==========================================
- Coverage   48.12%   48.09%   -0.03%     
==========================================
  Files         233      235       +2     
  Lines       29229    29281      +52     
==========================================
+ Hits        14066    14083      +17     
- Misses      14131    14168      +37     
+ Partials     1032     1030       -2     

☔ 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.

[jhrozek]

In general I think this is a very nice refactor, one question about concurrent access inline

[dmartinol]

@jhrozek @yrobla I reviewed the initial implementation again to:

• remove the Idle sync state (what does it mean? 🤔 ), replaced with Completed
• avoid syncing if there is any sync state requiring change (e.g. Failed) but before the scheduled requeue time (5m)
• avoid updating the sync state at all if sync is not needed

The above changes prevent continuous reconciliation loops in case of failures, properly store incremental attempt counter in the sync status and respect the scheduled retries.

Hopefully, the E2E tests will confirm that everything is fine

[jhrozek]

awesome work @dmartinol