fix: remove duplicate scrape_configs from prometheus.yml

[coder]

Summary

Fix two regressions introduced by #85's Prometheus upgrade in v0.7.4. v0.7.3 is unaffected.

• Remove the legacy serverFiles."prometheus.yml".scrape_configs: [] setting. It duplicates the new chart's scrape_configs key, causing Prometheus 3.13.2 to exit at startup. Chart-managed scraping remains disabled through scrapeConfigs.
• Replace the ignored --enable-feature=remote-write-receiver flag with --web.enable-remote-write-receiver so the collector can deliver metrics.
• Regenerate the compiled manifests and README, and sync chart metadata to v0.7.4.

CI coverage

Add Validate Prometheus config correctness: compile the configuration and alert rules, start the rendered Prometheus image with its rendered arguments, and require /-/ready to succeed within a bounded retry window. Print server logs and clean up afterward.

make test/prometheus depends on build. The shell check runs in PR/main CI, nightly builds, and against the packaged chart before publication. It validates startup and configuration loading; Kubernetes discovery and metrics ingestion are outside its scope.

Validation

• Ran the isolated CI command against v0.7.4 with only the test harness added: compilation passed, but readiness failed with field scrape_configs already set in type config.plain.
• The fixed source chart and packaged chart pass the readiness check.
• Separately verified that the receiver-flag fix permits writing and querying a metric.
• make lint and ShellCheck pass locally.