Use imbalanced-learn in the componentspecific model

[marco-c]

The dataset is small and not balanced (https://community-tc.services.mozilla.com/tasks/Woy2F_GaTXeJ29eCHoh1bg/runs/0/logs/public/logs/live.log):

                          pre       rec       spe        f1       geo       iba       sup
                 1       0.90      0.80      0.33      0.85      0.52      0.28        71
                 0       0.50      0.11      0.99      0.18      0.33      0.10         9
__NOT_CLASSIFIED__       0.00      0.00      0.81      0.00      0.00      0.00         0
       avg / total       0.86      0.72      0.41      0.78      0.50      0.26        80
╒════════════╤═════════════════╤═════════════════╤══════════════════╕
│            │   1 (Predicted) │   0 (Predicted) │   Not classified │
╞════════════╪═════════════════╪═════════════════╪══════════════════╡
│ 1 (Actual) │              57 │               1 │               13 │
├────────────┼─────────────────┼─────────────────┼──────────────────┤
│ 0 (Actual) │               6 │               1 │                2 │
╘════════════╧═════════════════╧═════════════════╧══════════════════╛

Originally posted by @suhaibmujahid in #5892 (comment)