Fix GPT-5.6 catalog pricing cache invalidation

[steipete]

Summary

• prefer an exact models.dev GPT-5.6 alias before falling back to the normalized Sol catalog entry
• fingerprint only priceable OpenAI/Anthropic cost data, while distinguishing missing and empty long-context blocks
• keep Pi caches stable across unrelated catalog metadata, but fully reprice Codex and Claude rows when relevant rates change

Follow-up correctness fixes for #2023 after its contributor branch merged while these review findings were being validated.

Tests

• swift test --no-parallel --filter 'CostUsagePricingTests|PiSessionCostScannerTests'
• make check
• make test — 595 selections in 50 groups; zero failures, retries, or timeouts
• two independent source/test reviews with no remaining findings

[clawsweeper]

ClawSweeper status: review started.

I am starting a fresh review of this pull request: Fix GPT-5.6 catalog pricing cache invalidation This is item 1/1 in the current shard. Shard 0/1.

This placeholder means the worker is alive and reading the current context. I will edit this same comment with the actual review when the claws are done clicking.

Crustacean status: shell secured, claws on keyboard, evidence pebbles being sorted.