Position: Added better collision detection for flip and fit, added visual tests for each and updated the unit tests to take the changes into account.

[kborchers]

In the process, I removed the rounding that was being done to the position since older jQuery couldn't handle the fractions. There was another pull for this same issue and I have merged the unit tests from that pull into this commit.

[jzaefferer]

Running the regular position unit tests with this applied, I get a bunch of failing tests. Most similar to the first one here: http://bassistance.de/i/319d21.png
Same results in Chrome 12, Firefox 5 and Safari 5.

[kborchers]

Yeah, I wasn't sure if it was just me not understanding the unit tests or not. I'll look at those a little later on today and see what I can figure out. Something I noticed yesterday though was now it seems like some browsers round fractional positions and others don't so I'm not sure how I should handle that if they are slightly different.

[inukshuk]

kborchers, I think the best approach for master is to never manipulate position in a way that could truncate pixel values (i.e, do not use Math.round, parseInt) and be extra careful when using division or subtraction. The thing is, if the browser returns fractions you want to keep them.

For 1.8-stable which may be distributed with older versions of jQuery you can test for support.fractions (defined in the position closure in this pull request: if false, the consensus was to use Math.round which may still lead to off-by-one errors in rare cases but at least gives you stable results.

[kborchers]

@jzaefferer This should be good to go now. All tests pass in Chrome, Firefox and Safari and all tests but the 5280 bug test pass in IE which I believe is a rounding thing. Not sure what to do about that other than @inukshuk's method for checking support of fractions.

[jzaefferer]

This breaks the flipping in the regular position demo. Set position to right (already the default in position_within), then drag to the right. It doesn't flip, instead just moves out of the screen (or the scrollable container).

[kborchers]

OK, I think that does it. Let me know if you find anything else. The one bug test fails in IE because IE rounds fractional values but otherwise all tests pass.

[kborchers]

I am going to close this pull request and submit another when I have this working better. I have noticed some oddities in fit tonight so I need to keep working.