Merge lp://qastaging/~liampwhite/inkscape/spfactory-gperf into lp://qastaging/~inkscape.dev/inkscape/trunk

I am skeptical about this.

1. Do we have any evidence that the strcmp chain is a performance problem?

2. At a minimum, this must be integrated with the build system, so that sp-factory.cpp is automatically rebuilt whenever sp-factory.gperf is modified.

3. I'd rather see something based on a trie initialized at runtime, since it would be more generally useful; for instance, we could reuse it for ID lookup.

1. No. strcmp is fast enough that most of the time you won't even notice the difference, especially for comparing small strings like these typenames.

2. That is the pain point of this merge proposal. IIRC win32 does not have gperf in its binaries. It might not be necessary but for pedanticism we would want to implement that (in both build systems, sigh.)

3. A trie has comparable or worse performance than a perfect hash table.

2. Gperf was added in the win32 devlibs rev. 55.But it's still missing in de win64 devlibs.