Fix persistence bug caused by untrustworthy HTMLFormElement.id getter

[botandrose]

Turns out we can't simply call .id on element nodes, because that property is unreliable on <form> and <fieldset> elements. It could return the id of the node as expected, OR (problematically) a child element that has a name attribute of id. The latter breaks persistence of the form element, which can have further consequences involving lost state.

This PR has two commits:

1. A failing test exposing the issue
2. A commit that fixes it, but at the cost of a 5% perf hit on my machine.

I'm marking this PR as WIP in the hopes I can find a faster way to resolve this.

For more information, see: https://developer.mozilla.org/en-US/docs/Web/API/HTMLFormElement#issues_with_naming_elements
Fixes #135

[botandrose]

Interestingly, it appears that the majority of the perf loss is simply in the function call itself... if I replace the implementation of the new id function with simply return node.id, the perf loss is about the same! I expected that JIT inlining would negate this, but apparently not.

Maybe a manual inline then? I don't love it, since we're accesing element ids all over the place... hmm.

[botandrose]

Okay, I got something working here with inlining the function. Perf is mostly unaffected now, and the bug is fixed, but code readability suffers. Hmm.