Optimize staticSearch to reduce build time and scale

[martindholmes]

Today I added a pile of very common terms to the stopwords list for lemdo-dev, and added a note to explain that on the lemdo-dev search page. That's one strategy for reducing the risk of out-of-memory errors during the build. This ticket will track other ideas as we come up with them and test them. Here are some:

• Aggressively seek out contexts which should not be indexed; this includes labels/captions, boilerplate headings, etc.
• Consider removing some search filters, such as Document history; has anyone ever used that?
• Consider reducing the number of kwic items and the kwic length. Both have been set to 10 up to now.

[martindholmes]

In rev 17446, I have reduced the kwic number and length to 5.

[martindholmes]

In rev 17452, I've implemented the exclusion list in jenkins/build_exclusions.txt.

[martindholmes]

In rev 17453, I've replaced the generation of the XML A-Z index with the plain-text version, so there will no longer be heavy XML or HTML versions of that index in the build.

[martindholmes]

in rev 17454, I've added more stopwords to the staticSearch list, and given more memory to the XML validation processes, which seem to be taking more time than they should (especially the Schematron).

[martindholmes]

I'm currently experimenting with running some components of the build in parallel.

[martindholmes]

I've switched to a partially-parallelized build target in rev 17456. On my desktop, that halved the build time, but it may not be so effective on Jenkins. We'll see.

[martindholmes]

I think our mission is accomplished here. Down from over 6.5 hours to 1.75.