Future: reintroduce Azure Document Intelligence OCR backend

[martsokha]

Status

Deferred — removed in #195 to clear the deck for the externalised inference architecture (#194). Reintroduce as opt-in when there's user demand.

Context

AzureDocaiBackend was a direct cloud OCR backend in nvisy-ocr/src/provider/azure_docai/, gated behind the azure-docai cargo feature (forwarded as microsoft through nvisy-engine/nvisy-server/nvisy-cli). It used Azure's async analyzeDocument POST + operationId polling flow.

The new architecture (#194) centres on externalised inference services (see nvisycom/inference). Cloud OCR backends carry provider-specific auth, polling, and parsing work that doesn't pay off without a user. This issue tracks reintroducing it when needed.

What this issue becomes

Add AzureDocaiBackend back to nvisy-ocr behind an azure-docai cargo feature.

Triggers to revisit

• A customer requests DocAI for self-hosted runtime
• We want DocAI as a fallback when the externalised Bento OCR is unavailable
• An Azure-resident deployment where calling DocAI directly is simpler than running a sidecar

Scope when reintroduced

• Restore nvisy-ocr/src/backend/azure_docai_backend.rs (new backend/-based layout)
• Restore the azure-docai feature on nvisy-ocr (with tokio/time for the poll loop)
• Restore the OcrBackend::AzureDocai { … } variant + OcrExtractor::from_config dispatch
• Restore feature forwarding nvisy-ocr/azure-docai → nvisy-engine/microsoft → nvisy-server/microsoft → nvisy-cli/microsoft
• Auth: Ocp-Apim-Subscription-Key header from env or config
• Poll: exponential backoff with a hard timeout cap; expose the cap on BentoNerParams-style config

Reference

The deleted code is preserved in git history; the last commit including it is the parent of the removal commit on #195.