xz supports parallel compression which obviously speeds up
the time taken to run the 'dist-xz' rule, but it also
speeds up *decompression* too, as the parallel compressor
creates different output which can be decompressed in parallel.

There's two options, I think, for automake:
1. Pass -T+1 to xz if xz is >= 5.4.0 (a runtime check
would be needed, but it's cheap to do) as this uses
1 thread, but uses the parallel compressor to do it,
so allows decompression in parallel by users;

2. Pass -T0 (or -Tn, n >= 1) to allow autodetection
of the number of available CPU cores (or n if specified)
which also then naturally uses the threaded compressor.