Code refactoring - Neighbour search algorithms

[LonelyCat124]

At the moment there is a lot of duplicated code in the neighbour search algorithms. The things that need doing are (at least):

• [x] Invoke functionality. The invoke functionality is independent from the dimensionality and periodicity but at the moment is duplicated amongst all the neighbour searches. This should be stripped out & moved into the utils folder to help with maintainability.
• [x] Privilege generating code. All the code that generates the privileges is reused over and over and it independent from almost everything, so it could easily be local read1_privs, read2_privs, ... = generate_symmetric_privileges(...)
• [x] Coherence generating code. As above, for now realistically this will always return exclusive anyway.
• [x] Multkernel generating code. Right now there's a version of the each "runner" for multikernel and single kernel version. Instead we can just have a combine_kernels(kernel_list) function globally to do this.

I want to add this in while I'm currently adding in reduction clauses if possible, once I've tested that they work correctly.

[LonelyCat124]

Its also possible task generated code is independent of periodicity - may be worth a future look.

Edit: This would require custom neighbour distance code/quote but could be doable at some point.

[LonelyCat124]

There is a start for the implementation for the last 3 bullet points now in the reduction_priv branch. Hopefully once done this will massively cut down on code replication.

[LonelyCat124]

Every thing is now done and tested for the firrst 3 bullet points.

[LonelyCat124]

This is done with improved performance also