Improve quickstart and default Wikidata settings (ad-freiburg#675)

* For large knowlege bases, the IndexBuilder opens thousands of temporary files at ones. This is not allowed on many linux sytems by default. The workaround for this problem is now documented in the quickstart.md
* The default batch size for the wikidata-settings.json is now 10 millions. This eliminates a common source for out-of-memory problems when building indices for Wikidata.

docs/quickstart.md

@@ -64,9 +64,12 @@ settings as explained in the previous paragraph). It takes about 20 hours on an
 AMD Ryzen 9 5900X. Note that the only difference is the basename (`wikidata`
 instead of `olympics`) and how the input files are piped into the
 `IndexBuilderMain` executable (using `bzcat` instead of `xzcat` and two files
-instead of one).
+instead of one). Also note the `ulimit -Sn 1048576`, which ensures that the
+operating system allows a sufficient number of open files (on some systems, the
+default is as low as `1024`, and for large datasets, QLever operates with more
+temporary files than that).
 
-        chmod o+w . && docker run -it --rm -v $QLEVER_HOME/qlever-indices/wikidata:/index --entrypoint bash qlever -c "cd /index && bzcat latest-all.ttl.bz2 latest-lexemes.ttl.bz2 | IndexBuilderMain -F ttl -f - -l -i wikidata -s wikidata.settings.json | tee wikidata.index-log.txt"
+        chmod o+w . && docker run -it --rm -v $QLEVER_HOME/qlever-indices/wikidata:/index --entrypoint bash qlever -c "cd /index && ulimit -Sn 1048576 && bzcat latest-all.ttl.bz2 latest-lexemes.ttl.bz2 | IndexBuilderMain -F ttl -f - -l -i wikidata -s wikidata.settings.json | tee wikidata.index-log.txt"
 
 ## Start the engine
 

examples/wikidata.settings.json

@@ -11,5 +11,5 @@
 	  "ignore-punctuation": true
   },
   "ascii-prefixes-only": true,
-  "num-triples-per-batch" : 50000000
+  "num-triples-per-batch" : 10000000
 }