WIP -- Accumulators (do not merge)

[stevelinton]

For anyone who is interested. Still lacks kernel code, documentation, tests, etc. etc.

[stevelinton]

Still lacks documentation and is built on the random-sources PR which breaks some tests.
If anyone is interested, my next plan is to try building versions of some of the stabchain functions on this and see if it makes a time or memory difference.

[stevelinton]

Replacing SCRSift by a version using an accumulator makes basically no difference to run-time and reduces garbage created by roughly half on one benchmark (size of a Sylow subgroup of S_350).

[fingolfin]

So what is the general idea here -- would we (potentially) also offer accumulators for things like integers, vectors, matrices?

I.e. is this meant to be a generic interface for "in-place" operations? That sounds potentially quite useful

[stevelinton]

@fingolfin See issue #550. We already have mutable vectors and matrices and suitable operations to modify them in place. This is for objects that do not have mutable versions.