Improve Validation Checks

[michaelmckinsey1]

Summary

Currently if a unification operation results in a duplicate index in the perfdata multi-index, users will hit an error during the _fill_perfdata function. By moving the validation check before this function, which checks for duplicates, we can provide a clearer error message, which results in faster debugging for us and the user.

Example

New

For cases without metadata_key

DuplicateIndexError: Duplicate index {798655681} found in DataFrame index.

For cases with metadata_key

DuplicateValueError: Different Thicket.metadata[metadata_key] may not contain duplicate values.
thickets[0].metadata['tuning'] and thickets[1].metadata['tuning'] have the same values: [{'block_128'}]

Old

ValueError: cannot handle a non-unique multi-index!

[michaelmckinsey1]

Depends on #125 unit test data