Convert floats representing integers to int in sample set serialization

[arcondello]

This does come with a performance hit. However it is relatively minor. Testing two edge cases on my system

import time

import dimod
import numpy as np

arr = np.full((100, 500000), 1, dtype=float)  # can be converted

t = time.perf_counter()
dimod.serialization.utils.serialize_ndarray(arr)
print(time.perf_counter() - t)

takes ~5.5 seconds on this branch and ~1 second on dimod 0.12.10.

import time

import dimod
import numpy as np

arr = np.full((100, 500000), 1.5, dtype=float)  # cannot be converted

t = time.perf_counter()
dimod.serialization.utils.serialize_ndarray(arr)
print(time.perf_counter() - t)

takes ~3.5 seconds on this branch and ~1 second on dimod 0.12.10.

For context, actually Json-serializing the resulting object at this scale takes ~27.8 seconds to dump to a file and ~6.7 seconds to dump to a string.

[codecov]

Codecov Report

Merging #1354 (3055f08) into main (e5a8e3b) will decrease coverage by 0.01%.
The diff coverage is 92.30%.

@@            Coverage Diff             @@
##             main    #1354      +/-   ##
==========================================
- Coverage   95.08%   95.07%   -0.01%     
==========================================
  Files          95       95              
  Lines        9902     9914      +12     
==========================================
+ Hits         9415     9426      +11     
- Misses        487      488       +1     

Files Changed	Coverage Δ
dimod/serialization/utils.py	88.70% <91.66%> (+0.70%)	⬆️
dimod/sampleset.py	91.43% <100.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[randomir]

LGTM

[randomir]

typo: missing ]

[randomir]

typos: received, possibly