Bug 207852

The grid track sizing algorithm handles intrinsic contributions from items that span a single track in a simpler way than when the span is greater than 1. In the former case, sizeTrackToFitNonSpanningItem only calculates the track sizing function of each track once per each item in that track. But in the latter case we not only calculate the track sizing function of multiple tracks per item, we also repeat this 5 times in order to handle the various TrackSizeComputationPhase. Calculating the used track sizing function from the raw one is not a very expensive operation when done once, but repeating it so many times has a performance impact. In Chromium I added a auto-grid-lots-of-spanning-data.html perf test. The changes that I plan to do will improve the performance of that test by 40% in WebKit.