Hotfix for LegacyMetric errors with 0 compute and 0 metric errors

[oelbert]

Description
The comparison in LegacyMetric if np.isnan(metric_err) or (metric_err > worst_metric_err): can fail if all metric errors are negative real numbers, leading to no "worst index" being assigned and causing a memory error. It also means large negative errors won't be considered as bad as small positive ones, which is undersirable. This PR fixes that by comparing the absolute value of the errors to find the worst.

Fixes #72

How Has This Been Tested?
Physics translate tests have been run with this

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• Any dependent changes have been merged and published in downstream modules
• New check tests, if applicable, are included

[FlorianDeconinck]

Good catch