Add support for Markdown in exercises

[Niols]

This PR is a small modification of the learnocaml_exercise.ml file that allows for support for Markdown in exercises description. It modifies the functions read_descr and read_descrs in such a way that:

• read_descr now has a new argument that is the list of extensions to try in order. I also fixed a bug that whould have happened if land was not empty in read_descrs.

• read_descrs now has a list of extensions to try that it can give to read_descr. I believe this list should at some point move somewhere else. Maybe in the descr record?

I had to add a small markdown_to_html function. I used the basic functions from Omd, which means that it might not support other features that you want to have in your flavour of Markdown. I believe there should be a module (Markdown?) that implements this flavour with an interface containing at least from_string and to_html.

[Niols]

Filename.extension already contains the dot. This is fixed in this PR. If the PR is not merged, this should be fixed anyway.

[Niols]

This function should probably go in a module that implements learn-ocaml's flavour of Markdown.

[yurug]

Yes. Do it if you want.

[Niols]

These lists should probably move at some point. Maybe the optionnal extensions should go in the descr record?

[yurug]

This PR is fine to me. @OCamlPro-Couderc I let you review it since you are the expert of this piece of code.

[AltGr]

This is great, thanks! Did'nt have time to get around to it yet :)
It would be best to plug in the markdown stuff that we use for the tutorials: it has, for example, support for mathjax snippets. I don't know how easy it would be, though.

[picdc]

This PR seems nice to me.
I was just thinking at one thing: we use "descr.html" as the key to represent the value descr. However, with the addition of internationalized description files and the support for markdown, this key is no longer relevant. It works for now since we convert everything into HTML and with one language by instance, but that's something we could improve in the future.

[Niols]

It would be best to plug in the markdown stuff that we use for the tutorials: it has, for example, support for mathjax snippets. I don't know how easy it would be, though.

@AltGr I do agree. There are a lot of things that could be extracted from https://github.com/ocaml-sf/learn-ocaml/blob/master/src/repo/learnocaml_tutorial_parser.ml. But the parse_tutorial functions are quite long and I don't pretend to understand exactly what everything does. It has to do with the Tutorial.text type found in https://github.com/ocaml-sf/learn-ocaml/blob/master/src/state/learnocaml_data.ml. This type and the functions converting from and to HTML and Markdown should probably have their own separate file (src/utils/richText.ml?).

I was just thinking at one thing: we use "descr.html" as the key to represent the value descr. However, with the addition of internationalized description files and the support for markdown, this key is no longer relevant. It works for now since we convert everything into HTML and with one language by instance, but that's something we could improve in the future.

@OCamlPro-Couderc I do agree too. The type 'a file is not generic enough to handle the descr file. I'm not sure if one should generalize it or agree that descr is a particular file.

[Niols]

There was a problem with the merge. It's because the PR was anterior to the "dunification" of learn-ocaml. And the Omd dependency lacks in this PR. One has to change https://github.com/ocaml-sf/learn-ocaml/blob/master/src/repo/dune#L6 and add "omd" as a dependency.