Support serde serialization of ensemble models

[dbrundu]

A fitted ensemble could not be persisted, even though the weak learners it wraps, like DecisionTree, already can (#449).

This PR adds the Serialize/Deserialize derives to EnsembleLearner<M> and AdaBoost<M, L>, following the pattern already used in linfa-trees and linfa-clustering, without manual impl. The serde feature needs to forward to linfa-trees/serde, which is required in order to get RandomForest serializable. Also, linfa-ensemble/serde is added to the serde cargo check in checking.yml.

Three unit tests are added to cross-check round-trip predictions and parameters between the original model and the serialized/reloaded one, by using serde-json. The prediction comparison for the RandomForest model was done by iterating over all the trees by hand, to get reproducible and meaningful comparison and not be dependent by the tree majority vote, which is dependent by the HashMap iteration order.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 77.89%. Comparing base (7fe5c86) to head (02c433f).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #450      +/-   ##
==========================================
+ Coverage   77.53%   77.89%   +0.36%     
==========================================
  Files         106      104       -2     
  Lines        7585     7524      -61     
==========================================
- Hits         5881     5861      -20     
+ Misses       1704     1663      -41     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.