Add JSX API

[Drup]

Follow up of #247

cc @Khady @jorisgio @ulrikstrid

[Drup]

Alright, the tests finally pass.

• The JSX doesn't depend on markup anymore. The common part of JSX and PPX are isolated in the tyxml-syntax package, and the parser-specific part is only in tyxml-ppx.
• The JSX supports namespaces ... reasonably. It's not perfect yet, but at least it tries to guess the correct one. On top, you can know specify a namespace by prefixing: <Svg.Line> .... </Svg.Line>. The first letter is turned lowercase (apparently, it's not syntactically valid to put a lowercase there ....).
• The list operator <> ..... </> is now recognized and supported
• Wrapping is supported

Current issues:

• All antiquotations must be single elements, unlike the PPX were they accept lists. List can only be spliced through the spread operator ...{foo}. Unfortunately, the spread operator is super limited in where you can put it, which makes it impossible to write several things that the PPX would accept. For instance [%html "<p>bar" elt1 "foo" elt2 "baz</p>"] should translate to <p> "bar" ...{elt1} "foo" ...{elt2} "baz" </p> which isn't allowed
• The code transformation is currently global. I will add an attribute to turn it on/off, and I'm also considering accepting <Tyxml.element /> to locally enable tyxml's jsx.

[Drup]

I added the toggle. The syntax is [@@@tyxml.jsx bool] in ocaml, and [@tyxml.jsx bool]; in reason.

@Khady I'm highly tempted to change the antiquotes so that in <foo> "bar" other </foo>, other must be a list of elements instead of a single one. A single element would be inserted by ([other]) (which is kinda verbose), and hopefully at some point, we can make reason accept [other] (without parens).
This would be much more consistent with the ppx, and more expressive.

[Khady]

Hmm I’m probably not the best person to ask as I’m not using jsx extensively. But I tend to believe that I prefer the current setup. - it is weird that “bar” is accepted and is a single element but others must be a list - most of the time we don’t have a list of elements from what I gathered in our code - when I write some HTML or use other template systems, each “element” is one element, not a list - I feel that using ...others once in a while is less heavy than using [other] all the time

[ulrikstrid]

I think keeping somewhat close to the react ppx would be preferable. I would rather have to spread a list than always wrap things with lists.

[Drup]

My issue is that the spread operator is only available if it's the only child. You can't write <> "foo" ...bar </>.

[Drup]

Everything's green, let's merge the current version! I kept the antiquotations as before.

@Khady @ulrikstrid I'll let you test this for a week. If someone's motivated to write a reason-inclined documentation, I would appreciate it as well ^^'