Major optimization in parser code

[bertfrees]

This is a long-standing issue that I finally had to tackle because due to our increasingly complex pipelines it had become impossible to execute our code without throwing a huge amount of memory at it.

The problem was that when parsing a XProc step declaration, every single import would result in parsing that file plus all imports of that file, leading to the same files possibly being parsed multiple times. And when you, like us, work with libraries that collect a lot of steps that in turn depend on other libraries, you are soon dealing with tens of thousands of files being parsed. This results in a huge memory consumption before the execution of the pipeline has even started.

Actually I'm a bit surprised that no one else has noticed the issue before. Nevertheless I think my fix will be interesting for others too.

In the end the solution that I came up with is pretty simple: I cache the PipelineLibrary objects. But in order to be able to do this I had to move some things around. I hope you like it.

[ndw]

Applied to version 1.2.5 for Saxon 9.9 and Saxon 10.x

[bertfrees]

Thanks!