[codex] Clarify optional OCR and Spark surfaces

[sidmohan0]

Summary

• Add a dedicated optional-surfaces docs page for OCR and Spark install profiles, limitations, and deferred-overhaul status.
• Keep README, docs index, SDK docs, CLI docs, and roadmap centered on lightweight text PII screening while preserving OCR/Spark references.
• Document exact extras for local OCR, URL-based OCR, Donut OCR, SparkService, and Spark PII UDF helpers.
• Add a runtime dependency safety test proving public OCR/Spark service references do not require optional OCR/Spark imports on the core path.

Stacking note

This branch is stacked on PR #141 (codex/dfpy-74-core-import-hardening) and therefore also includes PR #139 in the current GitHub diff. It targets dev so CI runs; after the earlier PRs merge, this should collapse to the incremental DFPY-75 docs/test changes.

Validation

• DATAFOG_NO_TELEMETRY=1 DO_NOT_TRACK=1 .venv312/bin/python -m pytest tests/test_runtime_dependency_safety.py tests/test_no_network_core.py -q
• .venv312/bin/python -m pre_commit run --files .gitignore README.md docs/index.rst docs/python-sdk.rst docs/cli.rst docs/optional-surfaces.rst docs/roadmap.rst tests/test_runtime_dependency_safety.py --show-diff-on-failure
• git diff --check
• DATAFOG_NO_TELEMETRY=1 DO_NOT_TRACK=1 .venv312/bin/python -m sphinx -b html docs docs/_build/html
• DATAFOG_NO_TELEMETRY=1 DO_NOT_TRACK=1 .venv312/bin/python -m pytest -m "not slow" -q

Refs DFPY-75