Don't do averaging where epsilon is uniform

[ChristopherHogan]

Implements the algorithm described in this comment. However, it doesn't seem to make any difference in the runtime of the simulation script in #764.
@stevengj @oskooi

[ChristopherHogan]

I must have made a mistake, because it also significantly changes the resulting epsilon (although it wasn't correct to begin with).

[stevengj]

Would be good to check that it is no longer messing up ε.

[ChristopherHogan]

The epsilon output is correct. Trying resolution 20 exceeds my available memory with the averaged material function, but at resolution 10 python/examples/ring.py and the version that uses a material function and the fallback averaging produces Harminv output that's the same up to 4 or 5 decimal places.

ring.py, resolution 10

harminv0:, frequency, imag. freq., Q, |amp|, amplitude, error
harminv0:, 0.11810150542150445, -0.0007317200176453875, 80.70129460278059, 0.003413351859500044, -0.0030498468152507087-0.0015327769962579412i, 1.0324565072157651e-05+0.0i
harminv0:, 0.1471626643261352, -0.00023257744313091665, 316.3734675750522, 0.028644166889685307, 0.019309509482952406-0.02115729520359148i, 7.310989427157562e-07+0.0i
harminv0:, 0.17524672525773047, -5.2271941035269695e-05, 1676.2982375141323, 0.007211538380799207, -0.0008129425141697651-0.007165571176563345i, 1.9077210115422416e-07+0.0i

Material function with averaging, resolution 10

harminv0:, frequency, imag. freq., Q, |amp|, amplitude, error
harminv0:, 0.11810148928382809, -0.0007316426211713271, 80.70982052327193, 0.0034121562439828577, -0.003048471978583548-0.001532849904309653i, 1.0408745501360815e-05+0.0i
harminv0:, 0.14716195963468734, -0.0002325660253292107, 316.3874848580549, 0.028641545695560697, 0.01932004384775394-0.021144125556564326i, 7.300059204362234e-07+0.0i
harminv0:, 0.17524619935908187, -5.242539743704413e-05, 1671.3864646379557, 0.007214293355171828, -0.0008172334836814178-0.007167855889150278i, 2.1592661960886496e-07+0.0i

I'll have to use AWS to do additional testing.