Better \kern in svg

[brucemiller]

Currently a \kern in svg adds the shift to the containing svg:g, but if there already were children, they get shifted as well. We really want to create a new svg:g in that case. Alas, there was clumsiness in other places dealing with svg:g, namely \hbox would wrap it's content in svg:g, but didn't actually close the correct container when done, so that got simplified as well. And finally, a few svg:g that got created by the driver, behind the scenes, needed to be marked as autoclose, so that they could be properly closed.

Thanks, @xworld21 for the excellent debugging, but I think this ends up being simpler, in the long run, than #2502.
check if this does what you expect.

Closes #2502

[dginev]

Since reviewing _autoclose effects is nearly impossible without context, should we add a test exercising this? Maybe in t/tikz which runs only with the "extra" test target.

[brucemiller]

It should show up in the xy test, but that doesn't seem to be there. There should be a simple version of it, though

[dginev]

The reason we have them in the "extra" tests is that we can't reproduce the exact same XML markup across texlives (as the internal tex attribute values change with the changes to pgf). So t/graphics isn't quite doable.

Edit: Ah, but that was true of pgf, not xy, fair.

[dginev]

Thank you for the test - it makes the difference nicely visible. Good to merge!

New PR version on the left, master on the right, using ar5iv.css

[brucemiller]

Yeah, xy is certainly more stable, and small tests are fast enough to include in the main tests (although the exhaustive test going around is too big & slow). So, now we've got a place to add other small xy tests, if we need.