Scalability break-down in `NNLSSolver` for more than 2 processors

[dreamer2368]

Scalability test using the unit test tests/test_NNLS in par-nnls branch shows a significant slow-down when more than 2 processors are used:

Number of processors	Time
1	382ms
2	249ms
3	10902ms
4	15296ms

This test result is from a m2-chip macbook.

This issue might be related to #269 .

[dreamer2368]

A scaling test on LC quartz (#271 ) shows a reasonable scaling result, suggesting this scaling result to be an anomaly from a local laptop.