Add support for processing instructions

[Delta-official]

Adds support for processing instructions in html5ever and changes tests to support ProcessingInstruction tokens (see: html5lib/html5lib-tests#199)

Fixes #789

[mrobinson]

A few comments here:

• Did you run tests for this change?
• Have you tested serialization? How does a processing instruction like <?one two?> serialize?
• It looks like you are only adding support for processing instructions in two insertion modes? The specification lists that they should be handled in many more (for instance in tables: https://html.spec.whatwg.org/multipage/parsing.html#parsing-main-intable).

[Delta-official]

A few comments here:

* Did you run tests for this change?

Yes. Some are failing (see below) but it's a result of html5lib-tests being outdated

Failing cases

failures:
tok: test3.test: <?
tok: test3.test: <? (exact errors)
tok: test3.test: <?A
tok: test3.test: <?B
tok: test3.test: <?B (exact errors)
tok: test3.test: <?Y (exact errors)
tok: test3.test: <?Z (exact errors)
tok: test3.test: <?A (exact errors)
tok: test3.test: <?a
tok: test3.test: <?a (exact errors)
tok: test3.test: <?Y
tok: test3.test: <?Z
tok: test3.test: <?z
tok: test3.test: <?z (exact errors)
tok: test3.test: <?b (exact errors)
tok: test3.test: <?y
tok: test3.test: <?b
tok: test3.test: <?y (exact errors)
tok: test2.test: Simili processing instruction
tok: test2.test: A bogus comment stops at >, even if preceded by two dashes
tok: test2.test: Simili processing instruction (exact errors)
tok: test2.test: A bogus comment stops at >, even if preceded by two dashes (exact errors)

* Have you tested serialization? How does a processing instruction like `<?one two?>` serialize?

Forgot about serialization, gonna add tests cases for that.

* It looks like you are only adding support for processing instructions in two insertion modes? The specification lists that they should be handled in many more (for instance in tables: https://html.spec.whatwg.org/multipage/parsing.html#parsing-main-intable).

Missed those. Mainly went by warnings to see what was broken since this is my first time contributing and the other modes had default branches

[mrobinson]

Looking better though I have a few minor comments. In addition you are still emitting the "bad char" error when the spec says to emit other kinds of parse errors.

[Delta-official]

In addition you are still emitting the "bad char" error when the spec says to emit other kinds of parse errors.

The rest of the tokenizer code uses bad_char_error, so I tried being consistent (unless there are plans to start emitting errors that the spec mandates, then i'll change it)

[mrobinson]

This one is getting close. Just a few minor things remaining:

[mrobinson]

@Delta-official Before this lands you will need to fix cargo test and remove the unreachable! branches for NodeData::ProcessingInstruction in rcdom/examples/print-rcdom.rs and rcdom/tests/html-tree-builder.rs. Thanks!

[Delta-official]

@mrobinson The current html5lib-tests submodule in the repo is outdated, which causes the test to fail; #795 updates it, so if you or someone else could review that it'd be appreciated