refactor: use train data if test data is empty for a cluster

[sathvikbhagavan]

As we sample train data and test data, it can happen that test data in one of the clusters is empty that leads to spurious CI failures.

This approach uses train data for obtaining best model in a cluster if there is no test data available as a fallback instead of erroring.

[codecov[bot]]

Codecov Report

Attention: 1 lines in your changes are missing coverage. Please review.

Comparison is base (4f069df) 65.41% compared to head (12a0529) 78.31%.

Files	Patch %	Lines
lib/SurrogatesMOE/src/SurrogatesMOE.jl	92.30%	1 Missing :warning:

Additional details and impacted files

```diff @@ Coverage Diff @@ ## master #466 +/- ## =========================================== + Coverage 65.41% 78.31% +12.89% =========================================== Files 23 23 Lines 3146 3159 +13 =========================================== + Hits 2058 2474 +416 + Misses 1088 685 -403 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.