Metrics: ensure ordinal encoder of classes is the same in real and synthetic datasets [type:bug]

[bvanbreugel]

Description

Fixes #250 (comment), by ensuring the ordinal encoders are fitted to the real and synthetic data concatenated.

Affected dependencies

None

Notes

• A quick fix that ensure the same encoder is used for real and synthetic data ordinal encodings. We may need to have a longer discussion on whether this is always preferential, and whether other optional datasets (e.g. augmented, reference) need to also be included in the encoder processing.
• Also note, Sklearn recommends against using LabelEncoder for features, I have not changed this to keep the PR minimal.
• When both datasets have the same classes (which was assumed by the previous code), functionality is identical

How has this been tested?

• Ran quick testing suit, giving "2159 passed, 13 skipped, 106 deselected, 2 xfailed, 2896 warnings"

Checklist

• I have followed the Contribution Guidelines and Code of Conduct
• I have commented my code following the van der Schaar Lab Styleguide
• I have labelled this PR with the relevant Type labels
• My changes are covered by tests

[bvanbreugel]

@robsdavis Could you approve this please?

[robsdavis]

Hi @bvanbreugel, Thanks for the PR - LGTM! I'm just doing a bug fix and then I'll merge this in next