Improve performance when printing large sample sets

[arcondello]

Currently the dimod.serialization.format.Formatter has poor performance when working with large sample sets.
Some obvious low hanging fruit is to exit early when the depth is exceeded

dimod/dimod/serialization/format.py

Line 380 in e72c86d

if depth is None:

but I imagine there are many other improvements.

[arcondello]

Another data point

import time

import dimod
import numpy as np

sampleset = dimod.SampleSet.from_samples(np.ones((91, 500000)), vartype='BINARY', energy=np.zeros(91))


def filter(sampleset, n=100):
    if len(sampleset.variables) > n:
        sampleset = dimod.keep_variables(sampleset, list(sampleset.variables)[:n])
    return sampleset

# no filter
t = time.perf_counter()
dimod.serialization.format.Formatter().format(sampleset)
print(time.perf_counter() - t)

# filter
t = time.perf_counter()
dimod.serialization.format.Formatter().format(filter(sampleset))
print(time.perf_counter() - t)

gives

21.41042103699965
0.05259135400046944

See also #1069 for the weird slicing syntax in the example.