fix: Separate logical replication byte-distance alerts

[philippemnoel]

What

Add dedicated logical WAL-position distance alerts matching MCC's thresholds: above 1 GiB for five minutes, or above 4 GiB for two minutes. Retain pod, database, and subscription identity, and explain that the received-versus-reported position gap does not measure committed apply progress.

Companion to #270: that PR changes the mixed lag rules to receipt-age alerts and fixes stopped-worker detection. Merge these together to retain separate byte-distance coverage; before #270 lands, the legacy mixed lag rules can also alert on the same gap. This PR does not change physical replication lag thresholds.

Tests

• Helm lint with monitoring and PrometheusRule enabled passed.
• Existing Prometheus-rule rendering checks and alert exclusions passed.
• Helm-rendered rule regression tests passed with vmalert-tool v1.148.0: Zero/negative distance, exact warning/critical boundaries, warning-only and critical gaps, same-named subscriptions in separate databases, and missing worker metrics.
• Added a Chainsaw test directory, automatically discovered by the existing CI workflow, to run the rendered rule fixtures.
• git diff --check passed.

Kubernetes integration remains for CI. No deployment was performed. Chart versions are generated by the publishing workflow.