Docs: the SQL EXPORT DATABASE settings that change in engine 26.12.1 (overwrite value, unknown names) #380

ArcadeData/arcadedb#11524 (merged as a7888faf80, in release 26.12.1) changes what the SQL EXPORT DATABASE ... WITH settings do, and the how-to page import-export.adoc (section "The SQL EXPORT DATABASE") says nothing about it yet.

What changes in the engine: WITH overwrite = false now keeps an existing file and refuses the export, the same as leaving overwrite out (before, the value was ignored and the file was replaced). The value must be true or false. A setting name the exporter does not know is refused with a hint (for example overwite), a file or d setting in a WITH clause no longer redirects the target, and an empty includeTypes, excludeTypes or includeRecords list is refused.

What stays true: db.export_database(...) goes through the Java API and already refuses an existing file unless overwrite=True (the page says so). I checked the six calls the bindings make against the merged head 446b031c85 on JDK 21 and 25 and all pass, so no code change is needed.

To do when the engine pin of the packages moves to 26.12.1: add two or three sentences to that section (the overwrite value, the refusal of unknown names, the empty lists), and say that on 26.11.1 and earlier WITH overwrite = false replaced an existing file.

Also in 26.12.1 (ArcadeData/arcadedb#11540, merged 692de48d1a, for #11539): EXPORT DATABASE has two new settings, compressionLevel (0 to 9, default 6) and compressionThreads (default automatic, half the cores up to 8; 0 keeps the old single-threaded gzip), on the SQL WITH clause, the Exporter API and the command line, for jsonl, graphml and graphson. The file is still one standard gzip stream. Add both to the same section of the how-to page, and say whether db.export_database(...) should pass them (the Python call has no compression_level or compression_threads argument yet: decide when the engine pin moves).