docs: correct the database connector page and document query_timeout

[marevol]

Summary

The 15.9 database connector page described behaviour the connector does not have,
and left out the parts most likely to trip an administrator up. This corrects it
and documents query_timeout, added in the fess-ds-db stack
(codelibs/fess-ds-db#19 … #28).

All seven languages: ja, en, de, es, fr, ko, zh-cn. 15.9 only.

Corrections

Was	Is
Download from Maven Central into WEB-INF/lib	maven.codelibs.org, into WEB-INF/plugin - where the admin console installs
Nothing about the JDBC driver beyond "place it in the classpath"	It is a separate download, and it has to reach the crawler process classpath: WEB-INF/lib or WEB-INF/env/crawler/lib
"BLOB, CLOB, NCLOB, byte array and binary stream columns are automatically passed through content extraction"	Only the binary forms are extracted. CLOB, NCLOB and character streams are read as text, and the MIME hints do not apply to them. Whether a BLOB arrives as java.sql.Blob or as a byte array is the driver's choice
url=url presented as a recipe	…with the warning that a missing url column makes the document URL the JDBC connection URL
Scripts see the result columns	Scripts see the whole parameter map, credentials included. Column label casing differs per database, and an unresolved label leaves the field unset rather than erroring
"Use environment variables / use the encryption features"	Both spelled out: FESS_ENV_ variables expand as ${...} (crawler.data.env.param.key.pattern), and a parameter matching app.encrypt.property.pattern is encrypted at rest. Plus a note that DEBUG logging masks these
Troubleshooting on ClassNotFoundException / Connection refused / SQLException	The messages the connector now emits, plus sections for missing parameters, per-row failures and "documents do not appear in search results"
fetch_size: "set to MIN_VALUE for MySQL streaming"	…and that other drivers reject the negative value, and what happens then
"Incremental crawling: narrow the SQL by hand"	…with what that costs. When a crawl finishes, Fess deletes the documents of that data store configuration that were not part of the crawl that just ran, so a filtered query leaves only the matching rows in the index. The section now says so and points at delete_old_docs=false

The delete_old_docs behaviour is in DataIndexHelper#deleteOldDocs(): it returns
early only when the parameter is literally false, so a partial crawl sweeps the
rest away by default.

New parameter

• query_timeout, with a note that stopping a job cannot interrupt a query
  blocked inside the driver - the stop is only checked between rows.

Verification

• Structural parity across all seven files: identical counts of .. note:: (7),
  .. warning:: (4), literal blocks (17) and list-table rows (18); section
  underlines checked with east-asian width taken into account.
• Every new parameter name, path and error message is present in all seven files,
  and no file still mentions repo1.maven.org, Maven Central,
  mysql-connector-j-8, ClassNotFoundException, No suitable driver or
  Connection refused.
• Each file parses under docutils with no message other than the expected
  Unknown interpreted text role "doc" from the Sphinx-only :doc: role in the
  "See Also" list.
• A full Sphinx build was not run: conf/conf.py fails to load under Sphinx 9.1
  with a unicodeescape error at line 228, which is unrelated to this change.

Note on scope

An earlier revision of this branch also documented incremental crawling through
${last_crawl_time}. That connector change (codelibs/fess-ds-db#29) was not
merged, so the parameters and the section describing them have been removed and
the manual-filter section restored - now carrying the delete_old_docs warning
it was missing.

One thing for the reviewer

The Spanish page carries no diacritics (Descripcion, Metodo,
contrasena). That is pre-existing and page-wide; #511, stacked on this branch,
fixes it across the seven affected Spanish pages.