Add redistribution benchmarks

Add timing benchmarks for redistributions between

1. g_lo (xyles, yxles or lexys) and le_lo

2. g_lo (xyles, yxles or lexys) and gf_lo

3. g_lo (xyles, yxles or lexys) and xxf_lo

4. xxf_lo and yxf_lo (see note)

In addition, this also add benchmarks for

• Repeated calls to the nonlinear term

• Repeated advances of the timestep

This requires the matching utils PR #3.

This is not a bugfix, but it would be convenient if this could go into master to make performance comparisons easier.

Note. The xxf_lo to yxf_lo test also has cases for xyles, yxles and lexys. These all do the same thing in the current data layout, but will be distinct in the future split_domains layout.