fix: re-register service when keepalive ping returns 404

[mortenoh]

Summary

• When the orchestrator loses a service entry (Redis TTL expiry, restart), keepalive pings return 404 forever with no recovery path. The only workaround was recreating the service container (docker compose up -d --force-recreate).
• Now _keepalive_loop detects 404 responses, waits a configurable grace period (default 30s, avoids thundering herd), then re-registers with the original parameters. On success the ping URL updates and keepalive resumes; on failure it retries next cycle.
• Adds RegistrationConfig pydantic model to carry re-registration parameters with type safety instead of a loose dict[str, Any].
• Exposes re_register_grace_period parameter on BaseServiceBuilder.with_registration().
• Bumps version to 0.9.0.

Test plan

• test_keepalive_reregisters_on_404 -- 404 triggers re-registration, ping_url updates
• test_keepalive_grace_period_respected -- re-registration waits for grace period
• test_keepalive_reregistration_failure_retries -- failed re-registration retries on next cycle
• test_keepalive_non_404_does_not_reregister -- 500 errors don't trigger re-registration
• test_keepalive_no_registration_kwargs_skips -- without config, 404 just logs (backward compat)
• All 375 existing tests pass, lint and type checks clean

[codecov]

Codecov Report

❌ Patch coverage is 86.48649% with 5 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/servicekit/api/registration.py	91.17%	2 Missing and 1 partial ⚠️
src/servicekit/api/service_builder.py	33.33%	2 Missing ⚠️

📢 Thoughts on this report? Let us know!