docs(15.9): add an OCR configuration page

[marevol]

Summary

Adds config/crawler-ocr.rst to the 15.9 tree in all seven languages and lists it after the thumbnail page in the crawler section of config/index.rst. The page covers:

• Scope: OCR runs when the tesseract command is installed and OCR is enabled. It applies to images, to images embedded in documents that Tika handles, and to PDFs that have no text layer.
• Scanned PDFs: when PDFBox gets no text, Fess extracts the PDF again with Tika and OCRs it. PDFs that already have some text are not OCR'd.
• Installation: installing Tesseract and its language data on Debian/Ubuntu and RHEL-family hosts.
• Settings: the new crawler.document.ocr.enabled, crawler.document.ocr.language and crawler.document.ocr.timeout settings, also settable as -Dfess.config.* options.
• Docker: the compose/tesseract recipe in docker-fess.
• Per-crawl override: the config.tika.tesseract.config parameter, which takes a classpath resource name.
• Operation: notes on CPU cost, the default excluded URLs of web configs, size limits and OCR accuracy.
• Upgrade note: a customized tika.xml from 15.8 or earlier still excludes TesseractOCRParser, so OCR stays off until that line is removed.
• Checking the setup: a short procedure, including the OCR is enabled and Tesseract OCR is not available lines in fess-crawler.log.

The documented behaviour comes from the companion fess and fess-crawler changes, and was checked on a live Fess 15.9.0-SNAPSHOT with Tesseract.

config/properties.rst is not regenerated here. Regenerating it from the current fess tree also picks up unrelated keys added since its last run, so it is better done as a separate sweep after the fess change is merged.

Checks

• tools/check_headings.py finds no heading problems in the new pages.
• The new pages and the updated index.rst were built with Sphinx in a scratch tree for every language, with no warnings from the new page.
• The de, es, fr, ko and zh-cn pages are translations of the en and ja pages.