fix(bind-rp): bound stuck named shutdowns

[bryanwieg]

Summary

• replace the unbounded rc.d stop wait with bounded RNDC, SIGTERM, and SIGKILL stages
• revalidate the captured named PID before every forced signal
• log graceful and escalated lifecycle events under the named syslog identity
• default BIND General Log to Informational so routine lifecycle records are visible
• bump os-bind-rp to 26.7_6

Incident behavior addressed

A named process could close port 53 during graceful shutdown and then hang indefinitely. The plugin waited forever, blocking the restart while process-only monitoring still reported named as alive. This change caps every command and wait, confirms the old PID is gone, and lets the existing restart chain continue only after a successful stop.

Verification

• exact commit 64508e6 built natively as os-bind-rp-26.7_6 and was installed on the HA backup node
• artifact SHA-256: 58cfe0493006a72ee6b211f917517cb58d1f111b02b0f4e9ce93672f682198c1
• normal package restart, named-checkconf, RNDC, DHCP watcher, UDP DNS, and TCP DNS checks passed
• forced SIGSTOP scenario escalated through SIGTERM and SIGKILL, removed the old PID, started a new PID, and restored UDP/TCP DNS while all CARP instances remained BACKUP
• canonical suite against the fixed source: 151 passed, 5 subtests passed
• independent standards, spec, code-simplifier, and test-suite-simplifier reviews: pass

Canonical tests and release-PR workflow support landed in #94.