Found bottleneck in core algo

[botandrose]

While profiling BARD Tracker on a particularly large DOM morph, I discovered that ~90% of the morph time was being spent in findBestMatch, specifically: if (cursor.contains(document.activeElement)) break; in findBestMatch. I've just commented it out in my local copy of idiomorph to fall back on focus restoration, but I want to explore how we might keep the functionality of native focus preservation without the slowdown. Maybe precomputing the active element's parents up to the morph root into a set, and looking for set inclusion of the cursor?