Make Azure live readiness observable

[jeremydixon22]

Summary

The first hosted Azure qualification authenticated and cleaned up correctly, but the disposable Function host never reached the first durable run. This change turns that opaque timeout into stage-specific, redacted evidence and removes avoidable waiting.

• retry Flex Consumption deployment immediately instead of sleeping for 90 seconds first
• require Azure to report indexed functions before obtaining a function key
• fall back from the default function key to the master key when necessary
• bound HTTP calls and record deployment, discovery, key, endpoint, and failure-stage evidence
• stop polling immediately when a durable run reaches a terminal failure state
• add a fast deterministic success/failure harness that proves receipt redaction and cleanup behavior

Validation

• Azure live script harness covers complete success and function-discovery failure
• actionlint and all shell syntax passed
• publication policy, change classifier, deterministic public export, and test-temp cleanup passed

The prior hosted run's cleanup receipt was green; no disposable Function, Storage account, or status container remains from it.