0.4: Add detailed counts to statistical comparison results

Track added, removed, filtered-out, and below-threshold statistics
separately in ScalarComparisonResult.  The 'different' DataFrame now
only contains rows present in both sides with differing values;
added/removed rows are exposed as dedicated attributes.

Statistical test verdicts report per-simulation breakdowns such as
'128 compared, 2 added, 1 removed, 5 filtered out, 3 different'
with the largest difference prefixed at the end.

compare_simulations statistical summary also shows ADDED/REMOVED
counts alongside IDENTICAL/DIFFERENT.

Fixes #14.