Managing memory footprint of building large test_trace_df

[dgkf]

In some cases with many unique tests and traces, reconstructing a test-trace data.frame from a large coverage object can consume an enormous amount of memory.

The objects themselves can be quite large, but this alone does not seem to account for the consumed memory which might be eaten up by intermediate objects that are not garbage collected while constructing the matrix.

Issue first identified with the rbmi package