Using `DataFlowTasks` for the parallelisation of LU factorisation

[maltezfaria]

It would be nice to use DataFlowTasks for the parallel LU factorisation. I am opening this issue to keep track of the changes and PRs needed to make that happen.

• Have a more robust set of benchmarks to judge progress/regression (Automate benchmarks #38)
• Profile and improve the current implementation of the LU factorisation #39
• Add @dspawn for the LU factorisation