Speed up wrapper generation

[kwabenantim]

Summary

Running PyChaste wrapper generation ( ~270 wrapped classes, ~1,300 headers) takes ~12 min.

Preliminary profiling results show:

• ~ 1/3 of the time is spent in the CastXML/pygccxml parse stage.
• ~ 2/3 of the time is spent in the Python wrapper generation code.

Options for potential speedup could involve:

• Optimising repetitive pathlib and regex operations.
• Tightening restrictions on source paths walked (e.g. build dirs).
• Optimising sort_classes.
• Optimising hierarchy_attribute searches.
• Parallelising the CastXML parse by splitting the header collection.